/***************************************************************************** Major portions of this software are copyrighted by the Medical College of Wisconsin, 1994-2000, and are released under the Gnu General Public License, Version 2. See the file README.Copyright for details. ******************************************************************************/ /*** Adapted from AlphaSim.c ***/ #include "mrilib.h" #ifdef USE_OMP #include #endif /**** C source codes to include, for compilation efficiency (we hope & pray) ***/ /*----------------------------------------------------------------------------*/ #include "zgaussian.c" /** Ziggurat Gaussian random number generator **/ /*----------------------------------------------------------------------------*/ #include "mri_radial_random_field.c" /** 3D FFT-based random field generator **/ /*----------------------------------------------------------------------------*/ #ifndef IND_T typedef unsigned short ind_t ; # define IND_T unsigned short # undef USE_UBYTE # define MAX_IND 65535u #endif #include "thd_Xdataset.c" /** reading .sdat datasets **/ /*----------------------------------------------------------------------------*/ #include #include #include #define MAX_NAME_LENGTH THD_MAX_NAME /* max. string length for file names */ #define MAX_CLUSTER_SIZE 99999 /* max. size of cluster for freq. table */ #undef ALLOW_LOHI #ifndef ALLOW_LOHI # define do_lohi 0 #else static int do_lohi = 0 ; #endif /*---------------------------------------------------------------------------*/ /* Global data */ static THD_3dim_dataset *mask_dset = NULL ; /* mask dataset */ static byte *mask_vol = NULL; /* mask volume */ static int mask_nvox = 0, mask_ngood = 0; /* number of good voxels in mask volume */ static int const max_cluster_size = MAX_CLUSTER_SIZE ; static int nx = 64 ; static int ny = 64 ; static int nz = 32 ; static int nxy ; static int nxyz ; static int nxyz1 ; static float dx = 3.5f ; static float dy = 3.5f ; static float dz = 3.5f ; static float fwhm_x = 0.0f ; static float fwhm_y = 0.0f ; static float fwhm_z = 0.0f ; static int niter = 10000 ; /*-- global variables for ACF simulation [30 Nov 2015] --*/ static int do_acf = 0 ; static float acf_a = 0.0f ; static float acf_b = 0.0f ; static float acf_c = 0.0f ; static float acf_parm[3] = {0.0f,0.0f,0.0f} ; static int acf_nxx=0 , acf_nyy=0 , acf_nzz=0 ; static MRI_IMAGE **acf_aim=NULL , **acf_bim=NULL ; static complex **acf_tar=NULL ; static int acf_ntar=0 ; static MRI_IMAGE *acf_wim=NULL ; static int do_classic = 0 ; /* does nothing (yet) */ static int ex_pad = 0 ; /* 12 May 2015 -- allow for padding the volume */ static int ey_pad = 0 ; static int ez_pad = 0 ; static int do_pad = 0 ; static int nx_pad ; static int ny_pad ; static int nz_pad ; static int nxy_pad ; static int nxyz_pad ; static int allow_padding = 1 ; /* on by default */ static float tdof = 0.0f ; /* 26 May 2015 - secret stuff */ static float sigmax , sigmay , sigmaz ; static int do_blur = 0 ; static int do_ssave = 0 ; /* 24 Apr 2014 */ static char *ssave_prefix = NULL ; static int ssave_index = 0 ; static THD_3dim_dataset *ssave_dset = NULL ; #define SSAVE_BLURRED 1 #define SSAVE_MASKED 2 #define SSAVE_ABSOLUTE 3 /* no longer used */ static int nodec = 0 ; static int do_niml = 0 ; static int do_1D = 1 ; static int do_ball = 0 ; static unsigned int gseed = 123456789 ; #define PMAX 0.5 static double pthr_init[9] = { 0.05 , 0.02 , 0.01, 0.005, 0.002, 0.001, 0.0005, 0.0002, 0.0001 } ; static double athr_init[4] = { 0.10 , 0.05 , 0.02 , 0.01 } ; static int npthr_lots = 29 ; static double pthr_lots[29] = { 0.10, 0.09, 0.08, 0.07, 0.06, 0.05, 0.04, 0.03, 0.02, 0.015, 0.01, 0.007, 0.005, 0.003, 0.002, 0.0015, 0.001, 0.0007, 0.0005, 0.0003, 0.0002, 0.00015, 0.0001, 0.00007, 0.00005, 0.00003, 0.00002, 0.000015, 0.00001 } ; static int npthr_mega = 38 ; static double pthr_mega[38] = { 0.100, 0.090, 0.080, 0.070, 0.060, 0.050, 0.045, 0.040, 0.035, 0.030, 0.025, 0.020, 0.015, 0.010, 0.009, 0.008, 0.007, 0.006, 0.005, 0.004, 0.003, 0.002, 0.001, 0.0009, 0.0008, 0.0007, 0.0006, 0.0005, 0.0004, 0.0003, 0.0002, 0.0001, 0.00007, 0.00005, 0.00003, 0.00002, 0.000015,0.00001 } ; static int nathr_lots = 10 ; static double athr_lots[10] = { 0.10, 0.09, .08, .07, .06, .05, .04, .03, .02, .01 } ; static int nathr_mega = 22 ; static double athr_mega[22] = { 0.100,0.095,0.090,0.085,0.080,0.075,0.070,0.065,0.060,0.055, 0.050,0.045,0.040,0.035,0.030,0.025,0.020,0.015,0.010,0.005,0.004,0.003 } ; static int force_mega = 0 ; /* 15 Dec 2021 */ static int did_mega = 0 ; #undef SET_MEGA /* 15 Dec 2021 */ #define SET_MEGA \ do{ npthr = npthr_mega ; \ pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; \ memcpy( pthr , pthr_mega , sizeof(double)*npthr ) ; \ nathr = nathr_mega ; \ athr = (double *)realloc(athr,sizeof(double)*nathr) ; \ memcpy( athr , athr_mega , sizeof(double)*nathr ) ; \ athr_sum_bot = 13 ; athr_sum_top = 31 ; \ if( niter < 30000 ) niter = 30000 ; \ did_mega = 1 ; \ } while(0) static int npthr = 9 ; static double *pthr = NULL ; static float *zthr_1sid = NULL ; static float *zthr_2sid = NULL ; static int nathr = 4 ; static double *athr = NULL ; /* the output: 2D table of cluster size threshold as function of per-voxel threshold and volumetric alpha (will be interpolated from simulation results) */ static float **clust_thresh_1sid_NN1 = NULL ; static float **clust_thresh_2sid_NN1 = NULL ; static float **clust_thresh_bsid_NN1 = NULL ; static float **clust_thresh_1sid_NN2 = NULL ; static float **clust_thresh_2sid_NN2 = NULL ; static float **clust_thresh_bsid_NN2 = NULL ; static float **clust_thresh_1sid_NN3 = NULL ; static float **clust_thresh_2sid_NN3 = NULL ; static float **clust_thresh_bsid_NN3 = NULL ; static int do_athr_sum = 0 ; /* 18 Dec 2015 */ static int athr_sum_bot=-1 , athr_sum_top=-1 ; /* max_table_1sid[nnn][ipthr][cc] is the count of how often the maximum cluster had exact size cc voxels (out of niter trials), at the pthr[ipthr] voxel-wise threshold p-value, for the nnn-th NN level (1-3), for 1-sided voxel-wise thresholding. Mutatis mutandis for max_table_2sid and max_table_bsid (bi-sided threshold) */ static int **max_table_1sid[4] , **max_table_2sid[4] , **max_table_bsid[4] ; /* cmx_table_1sid[nnn][ipthr] = max cluster size found over all simulations for 1-sided clustering with NN level nnn at voxel-wise threshold pthr[ipthr]. That is, it is the largest value of cc such that max_table_1sid[nnn][ipthr][cc] > 0. [23 Dec 2015] */ static int *cmx_table_1sid[4] , *cmx_table_2sid[4] , *cmx_table_bsid[4] ; /* csiz_1D_NN1[ipthr][iter] = max cluster size found at the ipthr-th threshold in the iter-th iteration. [23 Dec 2015] */ static int **csiz_1sid_NN1 , **csiz_1sid_NN2 , **csiz_1sid_NN3 ; static int **csiz_2sid_NN1 , **csiz_2sid_NN2 , **csiz_2sid_NN3 ; static int **csiz_bsid_NN1 , **csiz_bsid_NN2 , **csiz_bsid_NN3 ; #undef USE_SHAVE #ifdef USE_SHAVE #define SHAVE_MALLOC 1 #define SHAVE_MMAP 2 static int do_shave =0 ; size_t shave_siz=0 ; static int64_t shave_tot=0 ; static short *shave =NULL ; #endif static int verb = 1 ; static int nthr = 1 ; #undef DECLARE_ithr /* 30 Nov 2015 */ #ifdef USE_OMP # define DECLARE_ithr const int ithr=omp_get_thread_num() #else # define DECLARE_ithr const int ithr=0 #endif static int minmask = 128 ; /* 29 Mar 2011 */ static char *prefix = NULL ; static THD_3dim_dataset **inset = NULL ; /* 02 Feb 2016 */ static int num_inset = 0 ; static int *nval_inset = NULL ; static Xdataset *xinset = NULL ; /* 29 Aug 2016 */ #undef PSMALL #define PSMALL 1.e-15 static char *cmd_fname = "3dClustSim.cmd" ; /*----------------------------------------------------------------------------*/ /*! Threshold for upper tail probability of N(0,1) */ double zthresh( double pval ) { if( pval <= 0.0 ) pval = PSMALL ; else if( pval >= 1.0 ) pval = 1.0 - PSMALL ; return qginv(pval) ; } /*---------------------------------------------------------------------------*/ static int vsnn=0 ; static void vstep_reset(void){ vsnn=0; } static void vstep_print(void) { static char xx[10] = "0123456789" ; fprintf(stderr , "%c" , xx[vsnn%10] ) ; if( vsnn%10 == 9) fprintf(stderr,".") ; vsnn++ ; } /*---------------------------------------------------------------------------*/ void display_help_menu() { printf( "Usage: 3dClustSim [options]\n" "\n" "Program to estimate the probability of false positive (noise-only) clusters.\n" "An adaptation of Doug Ward's AlphaSim, streamlined for various purposes.\n" "\n" "-----------------------------------------------------------------------------\n" "This program has several different modes of operation, each one involving\n" "simulating noise-only random volumes, thresholding and clustering them,\n" "and counting statistics of how often data 'survives' these processes at\n" "various threshold combinations (per-voxel and cluster-size).\n" "\n" "OLDEST method = simulate noise volume assuming the spatial auto-correlation\n" " function (ACF) is given by a Gaussian-shaped function, where\n" " this shape is specified using the FWHM parameter. The FWHM\n" " parameter can be estimated by program 3dFWHMx.\n" " ** THIS METHOD IS NO LONGER RECOMMENDED **\n" "\n" "NEWER method = simulate noise volume assuming the ACF is given by a mixed-model\n" " of the form a*exp(-r*r/(2*b*b))+(1-a)*exp(-r/c), where a,b,c\n" " are 3 parameters giving the shape, and can also be estimated\n" " by program 3dFWHMx.\n" " ** THIS METHOD IS ACCEPTABLE **\n" "\n" "NEWEST method = program 3dttest++ simulates the noise volumes by randomizing\n" " and permuting input datasets, and sending those volumes into\n" " 3dClustSim directly. There is no built-in math model for the\n" " spatial ACF.\n" " ** THIS METHOD IS MOST ACCURATE AT CONTROLLING FALSE POSITIVE RATE **\n" " ** You invoke this method with the '-Clustsim' option in 3dttest++ **\n" "\n" "3dClustSim computes a cluster-size threshold for a given voxel-wise p-value\n" "threshold, such that the probability of anything surviving the dual thresholds\n" "is at some given level (specified by the '-athr' option).\n" "\n" "Note that this cluster-size threshold is the same for all brain regions.\n" "There is an implicit assumption that the noise spatial statistics are\n" "the same everywhere.\n" "\n" "Program 3dXClustSim introduces the idea of spatially variable cluster-size\n" "thresholds, which may be more useful in some cases. 3dXClustSim's method is\n" "invoked by using the '-ETAC' option in 3dttest++.\n" "-----------------------------------------------------------------------------\n" "\n" "**** NOTICE ****\n" "You should use the -acf method, NOT the -fwhm method, when determining\n" "cluster-size thresholds for FMRI data. The -acf method will give more\n" "accurate false positive rate (FPR) control.\n" "****************\n" "\n" "In particular, this program lets you run with multiple p-value thresholds\n" "(the '-pthr' option) and only outputs the cluster size threshold at chosen\n" "values of the alpha significance level (the '-athr' option).\n" "\n" "In addition, the program allows the output to be formatted for inclusion\n" "into an AFNI dataset's header, whence it can be used in the AFNI Clusterize\n" "interface to show approximate alpha values for the displayed clusters, where\n" "the per-voxel p-value is taken from the interactive threshold slider in the\n" "AFNI 'Define Overlay' control panel, and then the per-cluster alpha value\n" "is interpolated in this table from 3dClustSim. As you change the threshold\n" "slider, the per-voxel p-value (shown below the slider) changes, and then\n" "the interpolated alpha values are updated.\n" "\n" "************* IMPORTANT NOTE [Dec 2015] ***************************************\n" "A completely new method for estimating and using noise smoothness values is\n" "now available in 3dFWHMx and 3dClustSim. This method is implemented in the\n" "'-acf' options to both programs. 'ACF' stands for (spatial) AutoCorrelation\n" "Function, and it is estimated by calculating moments of differences out to\n" "a larger radius than before.\n" "\n" "Notably, real FMRI data does not actually have a Gaussian-shaped ACF, so the\n" "estimated ACF is then fit (in 3dFWHMx) to a mixed model (Gaussian plus\n" "mono-exponential) of the form\n" " ACF(r) = a * exp(-r*r/(2*b*b)) + (1-a)*exp(-r/c)\n" "where 'r' is the radius, and 'a', 'b', 'c' are the fitted parameters.\n" "The apparent FWHM from this model is usually somewhat larger in real data\n" "than the FWHM estimated from just the nearest-neighbor differences used\n" "in the 'classic' analysis.\n" "\n" "The longer tails provided by the mono-exponential are also significant.\n" "3dClustSim has also been modified to use the ACF model given above to generate\n" "noise random fields.\n" "\n" "**----------------------------------------------------------------------------**\n" "** The take-away (TL;DR or summary) message is that the 'classic' 3dFWHMx and **\n" "** 3dClustSim analysis, using a pure Gaussian ACF, is not very correct for **\n" "** FMRI data -- I cannot speak for PET or MEG data. **\n" #if 0 "** **\n" "** You should start using the '-acf' options in your own scripts. AFNI **\n" "** scripts from afni_proc.py are moving away from the 'classic' method. At **\n" "** some point in the future, '-acf' will become the default in both programs, **\n" "** and you will have to actively specify '-classic' to use the older method. **\n" #endif "**----------------------------------------------------------------------------**\n" "\n" "** ---------------------------------------------------------------------------**\n" "** IMPORTANT CHANGES -- February 2015 ******************************************\n" "** ---------------------------------------------------------------------------**\n" "** In the past, 3dClustSim did '1-sided' testing; that is, the random dataset\n" "** of Gaussian noise-only values is generated, and then it is thresholded on\n" "** the positive side so that the N(0,1) upper tail probability is pthr.\n" "**\n" "** NOW, 3dClustSim does 3 different types of thresholding:\n" "** 1-sided: as above\n" "** 2-sided: where positive and negative values above the threshold\n" "** are included, and then clustered together\n" "** (in this case, the threshold on the Gaussian values is)\n" "** (fixed so that the 1-sided tail probability is pthr/2.)\n" "** bi-sided: where positive values and negative values above the\n" "** threshold are clustered SEPARATELY (with the 2-sided threshold)\n" "** For high levels of smoothness, the results from bi-sided and 2-sided are\n" "** very similar -- since for smooth data, it is unlikely that large clusters of\n" "** positive and negative values will be next to each other. With high smoothness,\n" "** it is also true that the 2-sided results for 2*pthr will be similar to the\n" "** 1-sided results for pthr, for the same reason. Since 3dClustSim is meant to be\n" "** useful when the noise is NOT very smooth, we provide tables for all 3 cases.\n" "**\n" "** In particular, note that when the AFNI GUI threshold is set to a t-statistic,\n" "** 2-sided testing is what is usually appropriate -- in that case, the cluster\n" "** size thresholds tend to be smaller than the 1-sided case, which means that\n" "** more clusters tend to be significant than in the past.\n" "**\n" "** In addition, the 3 different NN approaches (NN=1, NN=2, NN=3) are ALL\n" "** always computed now. That is, 9 different tables are produced, each\n" "** of which has its proper place when combined with the AFNI Clusterize GUI.\n" "** The 3 different NN methods are:\n" "** 1 = Use first-nearest neighbor clustering\n" "** * above threshold voxels cluster together if faces touch\n" "** 2 = Use second-nearest neighbor clustering\n" "** * voxels cluster together if faces OR edges touch\n" "** 3 = Use third-nearest neighbor clustering\n" "** * voxels cluster together if faces OR edges OR corners touch\n" "** The clustering method only makes a difference at higher (less significant)\n" "** values of pthr. At small values of pthr (more significant), all three\n" "** clustering methods will give very similar results.\n" "**\n" "**** PLEASE NOTE that the NIML outputs from this new version are not named the\n" "**** same as those from the older version. Thus, any script that takes the NIML\n" "**** format tables and inserts them into an AFNI dataset header must be modified\n" "**** to match the new names. The 3drefit command fragment output at the end of\n" "**** this program (and echoed into file '3dClustSim.cmd') shows the new form\n" "**** of the names involved.\n" "**** -------------------------------------------------------------------------**\n" "**** SMOOTHING CHANGE -- May 2015 **********************************************\n" "** ---------------------------------------------------------------------------**\n" "** It was pointed out to me (by Anders Eklund and Tom Nichols) that smoothing\n" "** the simulated data over a finite volume introduces 2 artifacts, which might\n" "** be called 'edge effects'. To minimize these problems, this program now makes\n" "** extra-large (padded) simulated volumes before blurring, and then trims those\n" "** back down to the desired size, before continuing with the thresholding and\n" "** cluster-counting steps. To run 3dClustSim without this padding added, use\n" "** the new '-nopad' option.\n" #if 0 "** Also see the manuscript\n" "** A Eklund, T Nichols, M Andersson, and H Knutsson.\n" "** Empirically Investigating the Statistical Validity of SPM, FSL and AFNI\n" "** for Single Subject FMRI Analysis.\n" #endif "**** -------------------------------------------------------------------------**\n" "\n" "-------\n" "OPTIONS [at least 1 option is required, or you'll get this help message!]\n" "-------\n" " ******* Specify the volume over which the simulation will occur *******\n" "\n" " -----** (a) Directly give the spatial domain that will be used **-----\n" "\n" "-nxyz n1 n2 n3 = Size of 3D grid to use for simulation\n" " [default values = 64 64 32]\n" "-dxyz d1 d2 d3 = give all 3 voxel sizes at once\n" " [default values = 3.5 3.5 3.5]\n" "-BALL = inside the 3D grid, mask off points outside a ball\n" " at the center of the grid and touching the edges;\n" " this will keep about 1/2 the points in the 3D grid.\n" " [default = use all voxels in the 3D grid]\n" "\n" " -----** OR: (b) Specify the spatial domain using a dataset mask **-----\n" "\n" "-mask mset = Use the 0 sub-brick of dataset 'mset' as a mask\n" " to indicate which voxels to analyze (a sub-brick\n" " selector '[]' is allowed) \n" "\n" "-OKsmallmask = Allow small masks. Normally, a mask volume must have\n" " 128 or more nonzero voxels. However, IF you know what\n" " you are doing, and IF you are willing to live life on\n" " the edge of statistical catastrophe, then you can use\n" " this option to allow smaller masks -- in a sense, this\n" " is the 'consent form' for such strange shenanigans.\n" " * If you use this option, it must come BEFORE '-mask'.\n" " * Also read the 'CAUTION and CAVEAT' section, far below.\n" " -->>** This option is really only recommended for users who\n" " understand what they are doing. Misuse of this option\n" " could easily be construed as 'p-hacking'; for example,\n" " finding results, but your favorite cluster is too small\n" " to survive thresholding, so you post-hoc put a small mask\n" " down in that region. DON'T DO THIS!\n" "\n" " ** '-mask' means that '-nxyz' & '-dxyz' & '-BALL' will be ignored. **\n" "\n" " -----** OR: (c) Specify the spatial domain by directly giving simulated volumes **-----\n" "\n" "-inset iset [iset ...] = Read the 'iset' dataset(s) and use THESE volumes\n" " as the simulations to threshold and clusterize,\n" " [Feb 2016] rather than create the simulations internally.\n" " * For example, these datasets could come from\n" " 3dttest++ -toz -randomsign 1000 -setA ...\n" " * This can be combined with '-mask'.\n" " * Using '-inset' means that '-fwhm', '-acf', '-nopad',\n" " '-niter', and '-ssave' are ignored as meaningless.\n" "\n" " ---** the remaining options control how the simulation is done **---\n" "\n" "-fwhm s = Gaussian filter width (all 3 dimensions) in mm (non-negative)\n" " [default = 0.0 = no smoothing]\n" " * If you wish to set different smoothing amounts for each\n" " axis, you can instead use option\n" " -fwhmxyz sx sy sz\n" " to specify the three values separately.\n" " **** This option is no longer recommended, since FMRI data ****\n" " **** does not have a Gaussian-shaped spatial autocorrelation. ****\n" " **** Consider using '-acf' or '3dttest++ -Clustsim' instead. ****\n" "\n" "-acf a b c = Alternative to Gaussian filtering: use the spherical\n" " autocorrelation function parameters output by 3dFWHMx\n" " to do non-Gaussian (long-tailed) filtering.\n" " * Using '-acf' will make '-fwhm' pointless!\n" " * The 'a' parameter must be between 0 and 1.\n" " * The 'b' and 'c' parameters (scale radii) must be positive.\n" " * The spatial autocorrelation function is given by\n" " ACF(r) = a * exp(-r*r/(2*b*b)) + (1-a)*exp(-r/c)\n" " >>---------->>*** Combined with 3dFWHMx, the '-acf' method is now a\n" " recommended way to generate clustering statistics in AFNI!\n" " *** Alternative methods we also recommend:\n" " 3dttest++ with the -Clustsim and/or -ETAC options.\n" "\n" "-nopad = The program now [12 May 2015] adds 'padding' slices along\n" " each face to allow for edge effects of the smoothing process.\n" " If you want to turn this feature off, use the '-nopad' option.\n" " * For example, if you want to compare the 'old' (un-padded)\n" " results with the 'new' (padded) results.\n" " * '-nopad' has no effect when '-acf' is used, since that option\n" " automatically pads the volume when creating it (via FFTs) and\n" " then truncates it back to the desired size for clustering.\n" "\n" "-pthr p1 .. pn = list of uncorrected (per voxel) p-values at which to\n" " threshold the simulated images prior to clustering.\n" " [default = 0.05 0.02 0.01 0.005 0.002 0.001 0.0005 0.0002 0.0001]\n" "\n" "-athr a1 .. an = list of corrected (whole volume) alpha-values at which\n" " the simulation will print out the cluster size\n" " thresholds. For each 'p' and 'a', the smallest cluster\n" " size C(p,a) for which the probability of the 'p'-thresholded\n" " image having a noise-only cluster of size C is less than 'a'\n" " is the output (cf. the sample output, below)\n" " [default = 0.10 0.05 0.02 0.01]\n" "\n" " ** It is possible to use only ONE value in each of '-pthr' and **\n" " ** '-athr', and then you will get exactly one line of output **\n" " ** for each sided-ness and NN case. For example: **\n" " ** -pthr 0.001 -athr 0.05 **\n" "\n" " ** Both lists '-pthr' and '-athr' (of values between 0 and 0.2) **\n" " ** should be given in DESCENDING order. They will be sorted to be **\n" " ** that way in any case, and such is how the output will be given. **\n" "\n" " ** The list of values following '-pthr' or '-athr' can be replaced **\n" " ** with the single word 'LOTS', which will tell the program to use **\n" " ** a longer list of values for these probabilities [try it & see!] **\n" " ** (i.e., '-pthr LOTS' and/or '-athr LOTS' are legal options) **\n" "\n" "-LOTS = the same as using '-pthr LOTS -athr LOTS'\n" "-MEGA = adds even MORE values to the '-pthr' and '-athr' grids.\n" " * NOTE: you can also invoke '-MEGA' by setting environment\n" " variable AFNI_CLUSTSIM_MEGA to YES.\n" " * Doing this will over-ride any use of other options to set\n" " the '-pthr' and '-athr' lists!\n" "\n" "-iter n = number of Monte Carlo simulations [default = 10000]\n" "\n" "-nodec = normally, the program prints the cluster size threshold to\n" " 1 decimal place (e.g., 27.2). Of course, clusters only come\n" " with an integer number of voxels -- this fractional value\n" " is interpolated to give the desired alpha level. If you\n" " want no decimal places (so that 27.2 becomes 28), use '-nodec'.\n" "\n" "-seed S = random number seed [default seed = 123456789]\n" " * if seed=0, then program will quasi-randomize it\n" "\n" "-niml = Output the table in an XML/NIML format, rather than a .1D format.\n" " * This option is for use with other software programs;\n" " see the NOTES section below for details.\n" " * '-niml' also implicitly means '-LOTS'.\n" "\n" "-both = Output the table in XML/NIML format AND in .1D format.\n" " * You probably want to use '-prefix' with this option!\n" " Otherwise, everything is mixed together on stdout.\n" " * '-both' implies 'niml' which implies '-LOTS' (unless '-MEGA').\n" " So '-pthr' (if desired) should follow '-both'/'-niml'\n" "\n" "-prefix ppp = Write output for NN method #k to file 'ppp.NNk_Xsided.1D',\n" " for k=1, 2, 3, and for X=1sided, 2sided, bisided.\n" " * If '-prefix is not used, all results go to standard output.\n" " You will probably find this confusing.\n" " * If '-niml' is used, the filename is 'ppp.NNk_Xsided.niml'.\n" " To be clear, the 9 files that will be named\n" " ppp.NN1_1sided.niml ppp.NN1_2sided.niml ppp.NN1_bisided.niml\n" " ppp.NN2_1sided.niml ppp.NN2_2sided.niml ppp.NN2_bisided.niml\n" " ppp.NN3_1sided.niml ppp.NN3_2sided.niml ppp.NN3_bisided.niml\n" " * If '-niml' AND '-mask' are both used, then a compressed ASCII\n" " encoding of the mask volume is stored into file 'ppp.mask'.\n" " This string can be stored into a dataset header as an attribute\n" " with name AFNI_CLUSTSIM_MASK, and will be used in the AFNI\n" " Clusterize GUI, if present, to mask out above-threshold voxels\n" " before the clusterizing is done (which is how the mask is used\n" " here in 3dClustSim).\n" " * If the ASCII mask string is NOT stored into the statistics dataset\n" " header, then the Clusterize GUI will try to find the original\n" " mask dataset and use that instead. If that fails, then masking\n" " won't be done in the Clusterize process.\n" "\n" " -cmd ccc = Write command for putting results into a file's header to a file\n" " named 'ccc' instead of '3dClustSim.cmd'. This option is mostly\n" " to help with scripting, as in\n" " 3dClustSim -cmd XXX.cmd -prefix XXX.nii ...\n" " `cat XXX.cmd` XXX.nii\n" "\n" " -quiet = Don't print out the progress reports, etc.\n" " * Put this option first to silence most informational messages.\n" "\n" " -ssave:TYPE ssprefix = Save the un-thresholded generated random volumes into\n" " datasets ('-iter' of them). Here, 'TYPE' is one of these:\n" " * blurred == save the blurred 3D volume before masking\n" " * masked == save the blurred volume after masking\n" " The output datasets will actually get prefixes generated\n" " with the string 'ssprefix' being appended by a 6 digit\n" " integer (the iteration index), starting at 000000.\n" " (You can use SOMETHING.nii as a prefix; it will work OK.)\n" " N.B.: This option will slow the program down a lot,\n" " and was intended to help just one specific user.\n" "\n" "------\n" "NOTES:\n" "------\n" "* This program is like running AlphaSim once for each '-pthr' value and then\n" " extracting the relevant information from its 'Alpha' output column.\n" " ++ One reason for 3dClustSim to be used in place of AlphaSim is that it will\n" " be much faster than running AlphaSim multiple times.\n" " ++ Another reason is that the resulting table can be stored in an AFNI\n" " dataset's header, and used in the AFNI Clusterize GUI to see estimated\n" " cluster significance (alpha) levels.\n" "\n" "* To be clear, the C(p,alpha) thresholds that are calculated are for\n" " alpha = probability of a noise-only smooth random field, after masking\n" " and then thresholding at the given per-voxel p value, producing a cluster\n" " of voxels at least this big.\n" " ++ So if your cluster is larger than the C(p,0.01) threshold in size (say),\n" " then it is very unlikely that noise BY ITSELF produced this result.\n" " ++ This statement does not mean that ALL the voxels in the cluster are\n" " 'truly' active -- it means that at least SOME of them are (very probably)\n" " active. The statement of low probability (0.01 in this example) of a\n" " false positive result applies to the cluster as a whole, not to each\n" " voxel within the cluster.\n" "\n" "* To add the cluster simulation C(p,alpha) table to the header of an AFNI\n" " dataset, something like the following can be done [tcsh syntax]:\n" " set fx = ( `3dFWHMx -detrend time_series_dataset+orig` )\n" " 3dClustSim -mask mask+orig -acf $fx[5] $fx[6] $fx[7] -niml -prefix CStemp\n" " 3drefit -atrstring AFNI_CLUSTSIM_NN1_1sided file:CStemp.NN1_1sided.niml \\\n" " -atrstring AFNI_CLUSTSIM_MASK file:CStemp.mask \\\n" " statistics_dataset+orig\n" " rm -f CStemp.*\n" " AFNI's Clusterize GUI makes use of these attributes, if stored in a\n" " statistics dataset (e.g., something from 3dDeconvolve, 3dREMLfit, etc.).\n" "\n" " ** Nota Bene: afni_proc.py will automatically run 3dClustSim, and **\n" " *** put the results into the statistical results dataset for you. ***\n" " **** Another reason to use afni_proc.py for single-subject analyses! ****\n" "\n" "* 3dClustSim will print (to stderr) a 3drefit command fragment, similar\n" " to the one above, that you can use to add cluster tables to any\n" " relevant statistical datasets you have lolling about.\n" "\n" "* The C(p,alpha) table will be used in Clusterize to provide the cluster\n" " level alpha value when the AFNI GUI is set so that the Overlay threshold\n" " sub-brick is a statistical parameter (e.g., a t- or F-statistic), from which\n" " a per-voxel p-value can be calculated, so that Clusterize can interpolate\n" " in the C(p,alpha) table.\n" " ++ To be clear, the per-voxel p-value is taken from the AFNI GUI threshold\n" " slider (the p-value is shown beneath the slider), and then the C(p,alpha)\n" " table is inverse-interpolated to find the per-cluster alpha value for\n" " each different cluster size.\n" " ++ As you move the AFNI threshold slider, the per-voxel (uncorrected for\n" " multiple comparisons) p-value changes, the cluster sizes change (as fewer\n" " or more voxels are included), and so the reported per-cluster alpha\n" " values change for both reasons -- different p and different cluster size.\n" " ++ The alpha values reported are 'per-cluster', and are not themselves\n" " corrected for multiple comparisons ACROSS clusters. These alpha values\n" " are corrected for multiple comparisons WITHIN a cluster.\n" "\n" "* AFNI will use the NN1, NN2, NN3 tables as needed in its Clusterize\n" " interface if they are all stored in the statistics dataset header,\n" " depending on the NN level chosen in the Clusterize controller.\n" "\n" "* The blur estimates (provided to 3dClustSim via -acf) comes from using\n" " program 3dFWHMx.\n" "\n" "-------------------\n" "CAUTION and CAVEAT: [January 2011]\n" "-------------------\n" "* If you use a small ROI mask and also have a large blur, then it might happen\n" " that it is impossible to find a cluster size threshold C that works for a\n" " given (p,alpha) combination.\n" "\n" "* Generally speaking, C(p,alpha) gets smaller as p gets smaller and C(p,alpha)\n" " gets smaller as alpha gets larger. As a result, in a small mask with small p\n" " and large alpha, C(p,alpha) might shrink below 1. But clusters of size C\n" " less than 1 don't make any sense!\n" "\n" "* For example, suppose that for p=0.0005 that only 6%% of the simulations\n" " have ANY above-threshold voxels inside the ROI mask. In that case,\n" " C(p=0.0005,alpha=0.06) = 1. There is no smaller value of C where 10%%\n" " of the simulations have a cluster of size C or larger. Thus, it is\n" " impossible to find the cluster size threshold for the combination of\n" " p=0.0005 and alpha=0.10 in this case.\n" "\n" "* 3dClustSim will report a cluster size threshold of C=1 for such cases.\n" " It will also print (to stderr) a warning message for all the (p,alpha)\n" " combinations that had this problem.\n" #if 0 "\n" "* This issue arises because 3dClustSim reports C for a given alpha.\n" " In contrast, AlphaSim reports alpha for each given C, and leaves\n" " you to interpret the resulting table; it doesn't try to find C(p,alpha)\n" " for a given alpha, but finds alpha for various values of C.\n" "\n" "* If you wish to see this effect in action, the following commands\n" " can be used as a starting point:\n" "\n" "3dClustSim -nxyz 8 8 8 -dxyz 2 2 2 -fwhm 8 -niter 10000\n" "AlphaSim -nxyz 8 8 8 -dxyz 2 2 2 -fwhm 8 -niter 10000 -quiet -fast -pthr 0.0005\n" "\n" " From the 3dClustSim command above, you should get a warning message\n" " similar to this, right after the table (only the first 2 lines are shown):\n" "\n" "*+ WARNING: Simulation not effective for these cases:\n" " NN=1 pthr= 0.001000 alpha= 0.100 [max simulated alpha= 0.087]\n" #endif "\n" "-----------------------------\n" "---- RW Cox -- July 2010 ----\n" ) ; printf( "\n" "-------------\n" "SAMPLE OUTPUT from the command '3dClustSim -fwhm 7' [only the NN=1 1-sided results]\n" "-------------\n" "# 3dClustSim -fwhm 7\n" "# 1-sided thresholding\n" "# Grid: 64x64x32 3.50x3.50x3.50 mm^3 (131072 voxels)\n" "#\n" "# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels\n" "# -NN 1 | alpha = Prob(Cluster >= given size)\n" "# pthr | 0.100 0.050 0.020 0.010\n" "# ------ | ------ ------ ------ ------\n" " 0.050000 162.5 182.2 207.8 225.7\n" " 0.020000 64.3 71.0 80.5 88.5\n" " 0.010000 40.3 44.7 50.7 55.1\n" " 0.005000 28.0 31.2 34.9 38.1\n" " 0.002000 19.0 21.2 24.2 26.1\n" " 0.001000 14.6 16.3 18.9 20.5\n" " 0.000500 11.5 13.0 15.1 16.7\n" " 0.000200 8.7 10.0 11.6 12.8\n" " 0.000100 7.1 8.3 9.7 10.9\n" "\n" "e.g., for this sample volume, if the per-voxel p-value threshold is set\n" "at 0.005, then to keep the probability of getting a single noise-only\n" "cluster at 0.05 or less, the cluster size threshold should be 32 voxels\n" "(the next integer above 31.2).\n" "\n" "If you ran the same simulation with the '-nodec' option, then the last\n" "line above would be\n" " 0.000100 8 9 10 11\n" "If you set the per voxel p-value to 0.0001 (1e-4), and want the chance\n" "of a noise-only false-positive cluster to be 5%% or less, then the cluster\n" "size threshold would be 9 -- that is, you would keep all NN clusters with\n" "9 or more voxels.\n" "\n" "The header lines start with the '#' (commenting) character so that the result\n" "is a correctly formatted AFNI .1D file -- it can be used in 1dplot, etc.\n" ) ; PRINT_AFNI_OMP_USAGE("3dClustSim",NULL) ; PRINT_COMPILE_DATE ; exit(0); } /*---------------------------------------------------------------------------*/ /* Routine to initialize the input options (values are in global variables). */ void get_options( int argc , char **argv ) { char * ep; int nopt=1 , ii , have_pthr=0; ENTRY("get_options") ; pthr = (double *)malloc(sizeof(double)*npthr) ; memcpy( pthr , pthr_init , sizeof(double)*npthr ) ; athr = (double *)malloc(sizeof(double)*nathr) ; memcpy( athr , athr_init , sizeof(double)*nathr ) ; if( AFNI_yesenv("AFNI_CLUSTSIM_MEGA") ){ /* 15 Dec 2021 */ force_mega = 1 ; /* MEGA it is and MEGA it will be! */ do_niml = 1 ; INFO_message("NOTE: AFNI_CLUSTSIM_MEGA forces use of -MEGA") ; SET_MEGA ; } if( getenv("AFNI_RANDOM_SEEDVAL") != NULL ){ gseed = (unsigned int)AFNI_numenv("AFNI_RANDOM_SEEDVAL") ; if( gseed == 0 ) gseed = ((unsigned int)time(NULL)) + 17*(unsigned int)getpid() ; INFO_message("random seed set to %u from AFNI_RANDOM_SEEDVAL",gseed) ; } while( nopt < argc ){ /*----- -inset iii -----*/ if( strcmp(argv[nopt],"-inset") == 0 ){ /* 02 Feb 2016 */ int ii,nbad=0 ; THD_3dim_dataset *qset ; if( num_inset > 0 ) ERROR_exit("You can't use '-inset' more than once!") ; if( xinset != NULL ) ERROR_exit("You can't use '-inset' after '-insdat'!") ; if( ++nopt >= argc ) ERROR_exit("You need at least 1 argument after option '-inset'") ; for( ; nopt < argc && argv[nopt][0] != '-' ; nopt++ ){ /* ININFO_message("opening -inset '%s'",argv[nopt]) ; */ qset = THD_open_dataset(argv[nopt]) ; if( qset == NULL ){ ERROR_message("-inset '%s': failure to open dataset",argv[nopt]) ; nbad++ ; continue ; } for( ii=0 ; ii < DSET_NVALS(qset) ; ii++ ){ if( DSET_BRICK_TYPE(qset,ii) != MRI_float ){ ERROR_message("-inset '%s': all sub-bricks must be float :-(",argv[nopt]) ; nbad++ ; break ; } } if( num_inset > 0 && DSET_NVOX(qset) != DSET_NVOX(inset[0]) ){ ERROR_message("-inset '%s': grid size doesn't match other datasets",argv[nopt]) ; nbad++ ; } inset = (THD_3dim_dataset **)realloc( inset , sizeof(THD_3dim_dataset *)*(num_inset+1)) ; nval_inset = (int *) realloc( nval_inset, sizeof(int) *(num_inset+1)) ; inset[num_inset] = qset ; nval_inset[num_inset] = DSET_NVALS(qset) ; num_inset++ ; } if( num_inset == 0 ) ERROR_exit("no valid datasets opened after -inset :-(") ; if( nbad > 0 ) ERROR_exit("can't continue after above -inset problems") ; continue ; } /*----- -insdat mask sdata -----*/ if( strcasecmp(argv[nopt],"-insdat") == 0 ){ int nfile , ii ; if( xinset != NULL ) ERROR_exit("You can't use option '%s' more than once!",argv[nopt]) ; if( num_inset > 0 ) ERROR_exit("You can't use '-insdat' after '-inset'!") ; if( ++nopt >= argc-1 ) ERROR_exit("You need at least 2 arguments after option '%s'",argv[nopt-1]) ; for( ii=nopt ; ii < argc && argv[ii][0] != '-' ; ii++ ) ; /*nada*/ nfile = ii-nopt ; if( verb ) INFO_message("Loading %s datasets",argv[nopt-1]) ; xinset = open_Xdataset( argv[nopt], nfile-1,argv+(nopt+1) ) ; nopt += nfile ; continue ; } /*----- -nxyz n1 n2 n3 -----*/ if( strcmp(argv[nopt],"-nxyz") == 0 ){ nopt++ ; if( nopt+2 >= argc ) ERROR_exit("need 3 arguments after -nxyz") ; nx = (int)strtod(argv[nopt++],NULL); if( nx <= 0 ) ERROR_exit("illegal nx value") ; ny = (int)strtod(argv[nopt++],NULL); if( ny <= 0 ) ERROR_exit("illegal ny value") ; nz = (int)strtod(argv[nopt++],NULL); if( nz <= 0 ) ERROR_exit("illegal nz value") ; continue ; } /*----- -dxyz d1 d2 d3 -----*/ if( strcmp(argv[nopt],"-dxyz") == 0 ){ nopt++ ; if( nopt+2 >= argc ) ERROR_exit("need 3 arguments after -dxyz") ; dx = strtod(argv[nopt++],NULL); if( dx <= 0.0f ) ERROR_exit("illegal dx value") ; dy = strtod(argv[nopt++],NULL); if( dy <= 0.0f ) ERROR_exit("illegal dy value") ; dz = strtod(argv[nopt++],NULL); if( dz <= 0.0f ) ERROR_exit("illegal dz value") ; continue ; } /**** -OKsmallmask ****/ if( strcmp(argv[nopt],"-OKsmallmask") == 0 ){ /* 29 Mar 2011 */ minmask = 2 ; nopt++ ; continue ; } /**** -tdof dof ****/ if( strcmp(argv[nopt],"-tdof") == 0 ){ /* 26 May 2015 -- hidden option */ nopt++ ; if( nopt >= argc ) ERROR_exit("need argument after -tdof!") ; tdof = (float)strtod(argv[nopt],NULL) ; if( tdof > 0.0f && tdof < 4.0f ) ERROR_exit("illegal value after -tdof") ; nopt++ ; continue ; } /**** -mask mset ****/ if( strcmp(argv[nopt],"-mask") == 0 ){ if( mask_dset != NULL ) ERROR_exit("Can't use -mask twice!") ; nopt++ ; if( nopt >= argc ) ERROR_exit("need argument after -mask!") ; mask_dset = THD_open_dataset(argv[nopt]); if( mask_dset == NULL ) ERROR_exit("can't open -mask dataset!") ; mask_vol = THD_makemask( mask_dset , 0 , 1.0,0.0 ) ; if( mask_vol == NULL ) ERROR_exit("can't use -mask dataset!") ; mask_nvox = DSET_NVOX(mask_dset) ; DSET_unload(mask_dset) ; mask_ngood = THD_countmask( mask_nvox , mask_vol ) ; if( mask_ngood < minmask ){ if( minmask > 2 && mask_ngood > 2 ){ ERROR_message("-mask has only %d nonzero voxels; minimum allowed is %d.", mask_ngood , minmask ) ; ERROR_message("To run this simulation, please read the CAUTION and CAVEAT in -help,") ; ERROR_message("and then you can use the '-OKsmallmask' option if you so desire.") ; ERROR_exit("Cannot continue -- may we meet under happier circumstances!") ; } else if( mask_ngood == 0 ){ ERROR_exit("-mask has no nonzero voxels -- cannot use this at all :-(") ; } else { ERROR_exit("-mask has only %d nonzero voxel%s -- cannot use this :-(", mask_ngood , (mask_ngood > 1) ? "s" : "\0" ) ; } } if( verb ) INFO_message("%d voxels in mask (%.2f%% of total)", mask_ngood,100.0*mask_ngood/(double)mask_nvox) ; nopt++ ; continue ; } /*----- -fwhm s -----*/ if( strcasecmp(argv[nopt],"-fwhm") == 0 ){ nopt++; if( nopt >= argc ) ERROR_exit("need argument after -fwhm"); fwhm_x = (float)strtod(argv[nopt],&ep) ; if( fwhm_x < 0.0f || ep == argv[nopt] ) ERROR_exit("illegal value after -fwhm") ; WARNING_message( "Use of -fwhm is not advised; consider using -acf instead\n" " FMRI data does not have Gaussian-shaped smoothness!" ) ; fwhm_y = fwhm_z = fwhm_x ; nopt++; continue; } if( strncasecmp(argv[nopt],"-classic",6) == 0 ){ /* 01 Dec 2015 */ do_classic = 1 ; nopt++ ; continue ; } /*----- -acf a b c [30 Nov 2015] -----*/ if( strcasecmp(argv[nopt],"-acf") == 0 ){ nopt++; if( nopt+2 >= argc ) ERROR_exit("need 3 arguments after -acf"); acf_a = (float)strtod(argv[nopt++],&ep) ; if( acf_a < 0.0f || acf_a > 1.0f || ep == argv[nopt-1] ) ERROR_exit("-acf: 'a' value should be between 0 and 1 :-(") ; acf_b = (float)strtod(argv[nopt++],&ep) ; if( acf_b <= 0.0f || ep == argv[nopt-1] ) ERROR_exit("-acf: 'b' value should be positive :-(") ; acf_c = (float)strtod(argv[nopt++],&ep) ; if( acf_c <= 0.0f || ep == argv[nopt-1] ) ERROR_exit("-acf: 'c' value should be positive :-(") ; do_acf = 1 ; continue ; } /*----- -nopad -----*/ if( strcasecmp(argv[nopt],"-nopad") == 0 ){ /* 12 May 2015 */ allow_padding = 0 ; nopt++ ; continue ; } /*----- -fwhmxyz s s s -----*/ if( strcasecmp(argv[nopt],"-fwhmxyz") == 0 ){ nopt++; if(nopt+2 >= argc) ERROR_exit("need 3 arguments after -fwhmxyz"); fwhm_x = (float)strtod(argv[nopt++],&ep) ; if( fwhm_x < 0.0f || ep == argv[nopt-1] ) ERROR_exit("illegal x value after -fwhmxyz") ; fwhm_y = (float)strtod(argv[nopt++],&ep) ; if( fwhm_y < 0.0f || ep == argv[nopt-1] ) ERROR_exit("illegal y value after -fwhmxyz") ; fwhm_z = (float)strtod(argv[nopt++],&ep) ; if( fwhm_z < 0.0f || ep == argv[nopt-1] ) ERROR_exit("illegal z value after -fwhmxyz") ; WARNING_message( "Use of -fwhmxyz is not advised; consider using -acf instead\n" " FMRI data does not have Gaussian-shaped smoothness!" ) ; continue; } /*----- -iter n -----*/ if( strcmp(argv[nopt],"-iter") == 0 || strncmp(argv[nopt],"-nite",5) == 0 ){ nopt++; if( nopt >= argc ) ERROR_exit("need argument after %s",argv[nopt-1]); niter = (int)strtod(argv[nopt],NULL) ; if( niter < 2000 && !do_ssave ){ WARNING_message("-iter %d replaced by 2000",niter) ; niter = 2000 ; } else if( niter < 1 ){ ERROR_exit("-iter %s is illegal :-(",argv[nopt]) ; } nopt++; continue; } /*----- -seed S -----*/ if( strcasecmp(argv[nopt],"-seed") == 0 ){ nopt++; if( nopt >= argc ) ERROR_exit("need argument after %s",argv[nopt-1]); gseed = (unsigned int)strtol(argv[nopt],NULL,10) ; if( gseed == 0 ){ gseed = ((unsigned int)time(NULL)) + 17*(unsigned int)getpid() ; if( verb ) INFO_message("-seed 0 resets to %u",gseed) ; } nopt++; continue; } /*----- -LOTS -----*/ if( strcasecmp(argv[nopt],"-LOTS") == 0 ){ if( !did_mega ){ npthr = npthr_lots ; pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; memcpy( pthr , pthr_lots , sizeof(double)*npthr ) ; nathr = nathr_lots ; athr = (double *)realloc(athr,sizeof(double)*nathr) ; memcpy( athr , athr_lots , sizeof(double)*nathr ) ; athr_sum_bot = 10 ; athr_sum_top = 22 ; } nopt++ ; continue ; } /*----- -MEGA -----*/ if( strcasecmp(argv[nopt],"-MEGA") == 0 ){ /* 18 Dec 2015 */ if( !did_mega ){ npthr = npthr_mega ; pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; memcpy( pthr , pthr_mega , sizeof(double)*npthr ) ; nathr = nathr_mega ; athr = (double *)realloc(athr,sizeof(double)*nathr) ; memcpy( athr , athr_mega , sizeof(double)*nathr ) ; athr_sum_bot = 13 ; athr_sum_top = 31 ; if( niter < 30000 ) niter = 30000 ; } nopt++ ; continue ; } /*----- -sumup -----*/ if( strcasecmp(argv[nopt],"-sumup") == 0 ){ /* 18 Dec 2015 */ do_athr_sum++ ; nopt++ ; continue ; } #if 0 /*----- -2sided -----*/ if( strcasecmp(argv[nopt],"-2sided") == 0 ){ do_2sid = 1 ; nopt++ ; continue ; } if( strcasecmp(argv[nopt],"-1sided") == 0 ){ do_2sid = 0 ; nopt++ ; continue ; } #endif /*----- -pthr p -----*/ if( strcmp(argv[nopt],"-pthr") == 0 || strcmp(argv[nopt],"-pval") == 0 ){ nopt++; if( nopt >= argc ) ERROR_exit("need argument after %s",argv[nopt-1]); if( force_mega ){ WARNING_message("AFNI_CLUSTSIM_MEGA is YES -- skipping '-pthr' option") ; for( ii=nopt ; ii < argc && argv[ii][0] != '-' ; ii++ ) ; /*nada*/ continue ; } if( did_mega ) WARNING_message("-pthr option is over-riding earlier -MEGA option!") ; if( strcmp(argv[nopt],"LOTS") == 0 ){ /* built in big table of values */ npthr = npthr_lots ; pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; memcpy( pthr , pthr_lots , sizeof(double)*npthr ) ; nopt++ ; } else { /* a list of values */ for( ii=nopt ; ii < argc && argv[ii][0] != '-' ; ii++ ) ; /*nada*/ npthr = ii-nopt ; if( npthr <= 0 ) ERROR_exit("No positive values found after %s",argv[nopt-1]) ; pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; for( ii=0 ; ii < npthr ; ii++ ){ pthr[ii] = strtod(argv[nopt+ii],NULL) ; if( pthr[ii] <= 0.0 || pthr[ii] > PMAX ) ERROR_exit("value '%s' after '%s' is illegal!",argv[nopt+ii],argv[nopt-1]) ; } if( npthr > 1 ){ for( ii=0 ; ii < npthr ; ii++ ) pthr[ii] = -pthr[ii] ; /* sort into */ qsort_double( npthr , pthr ) ; /* descending */ for( ii=0 ; ii < npthr ; ii++ ) pthr[ii] = -pthr[ii] ; /* order */ for( ii=1 ; ii < npthr ; ii++ ){ if( pthr[ii] == pthr[ii-1] ) WARNING_message("duplicate value %g after '%s'",pthr[ii],argv[nopt-1]) ; } } nopt += npthr ; } have_pthr = 1 ; /* warn if -pthr is before -both/-niml */ continue ; } /*----- -athr p -----*/ if( strcmp(argv[nopt],"-athr") == 0 || strcmp(argv[nopt],"-aval") == 0 ){ nopt++; if( nopt >= argc ) ERROR_exit("need argument after %s",argv[nopt-1]); if( force_mega ){ WARNING_message("AFNI_CLUSTSIM_MEGA is YES -- skipping '-athr' option") ; for( ii=nopt ; ii < argc && argv[ii][0] != '-' ; ii++ ) ; /*nada*/ continue ; } if( did_mega ) WARNING_message("-athr option is over-riding earlier -MEGA option!") ; if( strcmp(argv[nopt],"LOTS") == 0 ){ /* built in big table of values */ nathr = nathr_lots ; athr = (double *)realloc(athr,sizeof(double)*nathr) ; memcpy( athr , athr_lots , sizeof(double)*nathr ) ; nopt++ ; } else { /* a list of values */ for( ii=nopt ; ii < argc && argv[ii][0] != '-' ; ii++ ) ; /*nada*/ nathr = ii-nopt ; if( nathr <= 0 ) ERROR_exit("No positive values found after %s",argv[nopt-1]) ; athr = (double *)realloc(athr,sizeof(double)*nathr) ; for( ii=0 ; ii < nathr ; ii++ ){ athr[ii] = strtod(argv[nopt+ii],NULL) ; if( athr[ii] <= 0.0 || athr[ii] > PMAX ) ERROR_exit("value '%s' after '%s' is illegal!",argv[nopt+ii],argv[nopt-1]) ; } if( nathr > 1 ){ for( ii=0 ; ii < nathr ; ii++ ) athr[ii] = -athr[ii] ; qsort_double( nathr , athr ) ; for( ii=0 ; ii < nathr ; ii++ ) athr[ii] = -athr[ii] ; for( ii=1 ; ii < nathr ; ii++ ){ if( athr[ii] == athr[ii-1] ) WARNING_message("duplicate value %g after '%s'",athr[ii],argv[nopt-1]) ; } } nopt += nathr ; } continue ; } /*---- -nodec ----*/ if( strcasecmp(argv[nopt],"-nodec") == 0 ){ nodec = 1 ; nopt++ ; continue ; } #ifdef ALLOW_LOHI if( strcasecmp(argv[nopt],"-alo") == 0 ){ /* 20 Jan 2011: debug stuff */ do_lohi = 1 ; nopt++ ; continue ; } #endif /*---- -niml ----*/ if( strcasecmp(argv[nopt],"-niml") == 0 || strcasecmp(argv[nopt],"-both") == 0 ){ /* let users know what to do with -pthr 19 May 2015 [rickr] */ if( have_pthr ) ERROR_exit("-both or -niml cannot follow -pthr, as they override it\n" " i.e. -pthr should be after -both or -niml") ; if( !did_mega ){ npthr = npthr_lots ; pthr = (double *)realloc(pthr,sizeof(double)*npthr) ; memcpy( pthr , pthr_lots , sizeof(double)*npthr ) ; nathr = nathr_lots ; athr = (double *)realloc(athr,sizeof(double)*nathr) ; memcpy( athr , athr_lots , sizeof(double)*nathr ) ; } do_niml = 1 ; do_1D = ( strcasecmp(argv[nopt],"-both") == 0 ) ; nopt++ ; continue ; } /*---- -BALL ----*/ if( strcasecmp(argv[nopt],"-BALL") == 0 ){ do_ball = 1 ; nopt++ ; continue ; } /*---- -NN ----*/ if( strcasecmp(argv[nopt],"-NN") == 0 ){ WARNING_message("-NN option is no longer supported! All NN cases are computed now :)") ; nopt++ ; if( isdigit(argv[nopt][0]) ) nopt++ ; continue ; } /*----- -prefix -----*/ if( strcmp(argv[nopt],"-prefix") == 0 ){ nopt++ ; if( nopt >= argc ) ERROR_exit("need argument after -prefix!") ; prefix = strdup(argv[nopt]) ; if( !THD_filename_ok(prefix) ) ERROR_exit("bad -prefix option!") ; nopt++ ; continue ; } /*----- -prefix -----*/ if( strcmp(argv[nopt],"-cmd") == 0 ){ nopt++ ; if( nopt >= argc ) ERROR_exit("need argument after -cmd!") ; cmd_fname = strdup(argv[nopt]) ; if( !THD_filename_ok(cmd_fname) ) ERROR_exit("bad -cmd option!") ; nopt++ ; continue ; } /*---- -quiet ----*/ if( strcasecmp(argv[nopt],"-quiet") == 0 ){ verb = 0 ; nopt++ ; continue ; } /*----- -ssave -----*/ if( strncasecmp(argv[nopt],"-ssave:",7) == 0 ){ if( nopt+1 >= argc ) ERROR_exit("need argument after %s",argv[nopt]) ; if( strcasecmp(argv[nopt]+7,"blurred") == 0 ){ do_ssave = SSAVE_BLURRED ; } else if( strcasecmp(argv[nopt]+7,"masked") == 0 ){ do_ssave = SSAVE_MASKED ; } else { ERROR_exit("this form of '-ssave' is unknown: '%s'",argv[nopt]) ; } ssave_prefix = strdup(argv[++nopt]) ; if( !THD_filename_ok(ssave_prefix) ) ERROR_exit("bad prefix after option '%s'",argv[nopt-1]) ; nopt++ ; continue ; } /*----- unknown option -----*/ ERROR_exit("3dClustSim -- unknown option '%s'",argv[nopt]) ; } /*------- finalize some simple setup stuff --------*/ if( fwhm_x > 0.0f ){ WARNING_message("**************************************************") ; WARNING_message( "I repeat:\n" " -fwhm uses a Gaussian-shaped autocorrelation function,\n" " which is not accurate for most FMRI data :(\n" " Consider using -acf instead, or 3dttest++ -Clustsim :)" ) ; WARNING_message("**************************************************") ; } if( do_athr_sum && (athr_sum_bot < 0 || athr_sum_top < 0) ){ /* 18 Dec 2015 */ do_athr_sum = 0 ; WARNING_message("-sumup canceled") ; } if( xinset != NULL && num_inset > 0 ) /* should not be possible */ ERROR_exit("-insdat and -inset are incompatible :-(") ; #define INSET_PRELOAD if( num_inset > 0 ){ /* 02 Feb 2016 */ int qq,nbad=0 ; nx = DSET_NX(inset[0]) ; ny = DSET_NY(inset[0]) ; nz = DSET_NZ(inset[0]) ; dx = fabsf(DSET_DX(inset[0])) ; dy = fabsf(DSET_DY(inset[0])) ; dz = fabsf(DSET_DZ(inset[0])) ; for( niter=qq=0 ; qq < num_inset ; qq++ ){ niter += nval_inset[qq] ; #ifdef INSET_PRELOAD ININFO_message("loading -inset '%s' with %d volumes",DSET_HEADNAME(inset[qq]),DSET_NVALS(inset[qq])) ; DSET_load(inset[qq]) ; if( !DSET_LOADED(inset[qq]) ){ ERROR_message("Can't load dataset -inset '%s'",DSET_HEADNAME(inset[qq])) ; nbad ++ ; } #endif } if( nbad > 0 ) ERROR_exit("Can't continue without all -inset datasets :-(") ; if( niter < 100 ) WARNING_message("-inset has only %d volumes total (= new '-niter' value)",niter) ; else if( verb ) INFO_message("-inset had %d volumes = new '-niter' value",niter) ; if( mask_dset != NULL ){ if( nx != DSET_NX(mask_dset) || ny != DSET_NY(mask_dset) || nz != DSET_NZ(mask_dset) ) ERROR_exit("-mask and -inset don't match in grid dimensions :-(") ; } if( do_ssave ){ WARNING_message("-inset turns off -ssave"); do_ssave = 0; } if( do_acf ){ WARNING_message("-inset turns off -acf") ; do_acf = 0 ; } } else if( xinset != NULL ){ /*--- 29 Aug 2016 ---*/ mask_dset = xinset->mask_dset ; mask_ngood = xinset->ngood ; mask_vol = xinset->mask_vol ; nx = DSET_NX(mask_dset) ; ny = DSET_NY(mask_dset) ; nz = DSET_NZ(mask_dset) ; dx = fabsf(DSET_DX(mask_dset)) ; dy = fabsf(DSET_DY(mask_dset)) ; dz = fabsf(DSET_DZ(mask_dset)) ; niter = xinset->nvtot ; if( niter < 100 ) WARNING_message("-insdat has only %d volumes total (= new '-niter' value)",niter) ; else if( verb ) INFO_message("-insdat had %d volumes = new '-niter' value",niter) ; if( do_ssave ){ WARNING_message("-insdat turns off -ssave"); do_ssave = 0; } if( do_acf ){ WARNING_message("-insdat turns off -acf") ; do_acf = 0; } } else if( mask_dset != NULL ){ nx = DSET_NX(mask_dset) ; ny = DSET_NY(mask_dset) ; nz = DSET_NZ(mask_dset) ; dx = fabsf(DSET_DX(mask_dset)) ; dy = fabsf(DSET_DY(mask_dset)) ; dz = fabsf(DSET_DZ(mask_dset)) ; if( do_ssave > 0 ){ /* 24 Apr 2014 */ ssave_dset = EDIT_empty_copy( mask_dset ) ; EDIT_dset_items( ssave_dset , ADN_nvals , 1 , ADN_prefix , "RandomJunk" , ADN_none ) ; } } if( do_acf ){ /* 30 Nov 2015 */ float val ; int_triple ijk ; if( fwhm_x > 0.0f ) WARNING_message("-acf option ==> any -fwhm options are ignored!") ; if( nx < 4 || ny < 4 || nz < 4 ) ERROR_exit("-acf option: minimum grid is 4x4x4, but you have %dx%dx%d",nx,ny,nz) ; fwhm_x = fwhm_y = fwhm_z = 0.0f ; acf_parm[0] = acf_a ; acf_parm[1] = acf_b ; acf_parm[2] = acf_c ; val = 2.0f * rfunc_inv(0.5f,acf_parm) ; ijk = get_random_field_size( nx,ny,nz , dx,dy,dz , acf_parm ) ; acf_nxx = ijk.i ; acf_nyy = ijk.j ; acf_nzz = ijk.k ; INFO_message("ACF(%.2f,%.2f,%.2f) => FWHM=%.2f => %dx%dx%d pads to %dx%dx%d", acf_a,acf_b,acf_c,val, nx,ny,nz , acf_nxx,acf_nyy,acf_nzz ) ; acf_wim = make_radial_weight( acf_nxx,acf_nyy,acf_nzz , dx,dy,dz , acf_parm ) ; acf_ntar = acf_nxx+acf_nyy+acf_nzz ; } if( do_ssave > 0 && ssave_dset == NULL ){ /* 24 Apr 2014 */ char gstr[128] ; float xorg,yorg,zorg ; xorg = -0.5*dx*(nx-1); yorg = -0.5*dy*(ny-1); zorg = -0.5*dz*(nz-1); sprintf(gstr,"RAI:D:%d,%f,%f,%d,%f,%f,%d,%f,%f", nx,xorg,dx , ny,yorg,dy , nz,zorg,dz ) ; ssave_dset = EDIT_geometry_constructor(gstr,"RandomJunque") ; } nxy = nx*ny ; nxyz = nxy*nz ; nxyz1 = nxyz - nxy ; if( nxyz < 256 ) ERROR_exit("Only %d voxels in simulation?! Need at least 256.",nxyz) ; if( mask_vol == NULL && do_ball ){ float xq,yq,zq , nx2,ny2,nz2 , nxq,nyq,nzq ; int ii,jj,kk ; nx2 = 0.5f*(nx-1) ; nxq = nx2*nx2 + 0.501f ; ny2 = 0.5f*(ny-1) ; nyq = ny2*ny2 + 0.501f ; nz2 = 0.5f*(nz-1) ; nzq = nz2*nz2 + 0.501f ; mask_vol = (byte *)malloc(sizeof(byte)*nxyz) ; for( kk=0 ; kk < nz ; kk++ ){ zq = kk - nz2 ; zq = (zq*zq) / nzq ; for( jj=0 ; jj < ny ; jj++ ){ yq = jj - ny2 ; yq = (yq*yq) / nyq ; for( ii=0 ; ii < nx ; ii++ ){ xq = ii - nx2 ; xq = (xq*xq) / nxq ; mask_vol[ii+jj*nx+kk*nxy] = ((xq+yq+zq) <= 1.0f) ; }}} mask_nvox = nxyz ; mask_ngood = THD_countmask( mask_nvox , mask_vol ) ; if( verb ) INFO_message("%d voxels in BALL mask (%.1f%% of total)", mask_ngood,100.0*mask_ngood/(double)mask_nvox) ; } if( mask_ngood == 0 ) mask_ngood = nxyz ; srand48(gseed) ; /* not really needed */ /*-- z-score thresholds for the various p-values --*/ zthr_1sid = (float *)malloc(sizeof(float)*npthr) ; zthr_2sid = (float *)malloc(sizeof(float)*npthr) ; for( ii=0 ; ii < npthr ; ii++ ){ zthr_1sid[ii] = (float)zthresh( pthr[ii] ) ; zthr_2sid[ii] = (float)zthresh( 0.5*pthr[ii] ) ; } /*-- check niter vs smallest alpha level --*/ if( niter*athr[nathr-1] < 10.0f ) WARNING_message( "Smallest athr=%g ==> should have at least %d iterations instead of %d" , athr[nathr-1] , (int)(10.0f/athr[nathr-1])+1 , niter ) ; /*-- blurring stuff --*/ sigmax = FWHM_TO_SIGMA(fwhm_x) ; if( nx < 2 ) sigmax = 0.0f ; sigmay = FWHM_TO_SIGMA(fwhm_y) ; if( ny < 2 ) sigmay = 0.0f ; sigmaz = FWHM_TO_SIGMA(fwhm_z) ; if( nz < 2 ) sigmaz = 0.0f ; do_blur = (sigmax > 0.0f || sigmay > 0.0f || sigmaz > 0.0f ) ; nx_pad = nx ; ny_pad = ny ; nz_pad = nz ; if( num_inset == 0 ){ if( do_acf ){ /* 30 Nov 2015 */ ex_pad = (acf_nxx-nx)/2 ; ey_pad = (acf_nyy-ny)/2 ; ez_pad = (acf_nzz-nz)/2 ; nx_pad = acf_nxx ; ny_pad = acf_nyy ; nz_pad = acf_nzz ; } else if( do_blur && allow_padding ){ /* 12 May 2015 */ ex_pad = (int)rintf(1.666f*fwhm_x/dx) ; ey_pad = (int)rintf(1.666f*fwhm_y/dy) ; ez_pad = (int)rintf(1.666f*fwhm_z/dz) ; do_pad = (ex_pad > 0) || (ey_pad > 0) || (ez_pad > 0) ; nx_pad = nx + 2*ex_pad ; ny_pad = ny + 2*ey_pad ; nz_pad = nz + 2*ez_pad ; if( do_pad ) INFO_message( "Padding by %d x %d x %d slices to allow for edge effects of blurring" , ex_pad,ey_pad,ez_pad) ; } } nxy_pad = nx_pad * ny_pad ; nxyz_pad = nxy_pad * nz_pad ; EXRETURN ; } /*---------------------------------------------------------------------------*/ void ssave_dataset( float *fim ) /* 24 Apr 2014 */ { #pragma omp critical { char *spr , sbuf[16] ; sprintf(sbuf,"_%06d",ssave_index) ; ssave_index++ ; spr = modify_afni_prefix(ssave_prefix,NULL,sbuf) ; EDIT_substitute_brick(ssave_dset,0,MRI_float,fim) ; EDIT_dset_items(ssave_dset,ADN_prefix,spr,ADN_none) ; ssave_dset->idcode = MCW_new_idcode() ; DSET_write(ssave_dset) ; DSET_NULL_ARRAY(ssave_dset,0) ; } return ; } /*---------------------------------------------------------------------------*/ /* Create the "functional" image, from the xinset struct [29 Aug 2016] */ void generate_fim_insdat( float *fim , int ival ) { load_from_Xdataset( xinset , ival , fim ) ; return ; } /*---------------------------------------------------------------------------*/ /* Create the "functional" image, from the inset dataset [02 Feb 2016] */ void generate_fim_inset( float *fim , int ival ) { if( ival < 0 || ival >= niter ){ /* should not be possible */ ERROR_message("inset[%d] == out of range!",ival) ; memset( fim , 0 , sizeof(float)*nxyz ) ; } else { float *bar=NULL ; int ii,qq,qval ; for( qval=ival,qq=0 ; qq < num_inset ; qq++ ){ /* find which */ if( qval < nval_inset[qq] ) break ; /* inset[qq] to use */ qval -= nval_inset[qq] ; } if( qq == num_inset ){ /* should not be possible */ ERROR_message("inset[%d] == array overflow !!",ival) ; } else { bar = DSET_ARRAY(inset[qq],qval) ; if( bar == NULL ){ ININFO_message("loading -inset '%s' with %d volumes",DSET_HEADNAME(inset[qq]),DSET_NVALS(inset[qq])) ; DSET_load(inset[qq]) ; bar = DSET_ARRAY(inset[qq],qval) ; } } if( bar != NULL ){ for( ii=0 ; ii < nxyz ; ii++ ) fim[ii] = bar[ii] ; /* copy data */ DSET_unload_one(inset[qq],qval) ; } else { ERROR_message("inset[%d] == NULL :-(",ival) ; memset( fim , 0 , sizeof(float)*nxyz ) ; } } } /*---------------------------------------------------------------------------*/ /* Create the "functional" image, the ACF way [30 Nov 2015]. */ void generate_fim_acf( float *fim , unsigned short xran[] ) { DECLARE_ithr ; int ii,jj,kk,pp,qq ; float *afim ; int do_a ; /* the function generates images in pairs, but 3dClustSim only consumes them one at a time -- thus the artifice of the 'aim' and 'bim' images, which are created together, then aim is consumed right away and bim is consumed on the next call here. */ if( acf_aim[ithr] == NULL && acf_bim[ithr] == NULL ){ /* need new aim & bim */ MRI_IMARR *imar ; imar = make_radial_random_field( acf_nxx , acf_nyy , acf_nzz , acf_wim , acf_tar[ithr] , xran ) ; acf_aim[ithr] = IMARR_SUBIM(imar,0) ; acf_bim[ithr] = IMARR_SUBIM(imar,1) ; FREE_IMARR(imar) ; } do_a = (acf_aim[ithr] != NULL) ; afim = (do_a) ? MRI_FLOAT_PTR(acf_aim[ithr]) : MRI_FLOAT_PTR(acf_bim[ithr]) ; /* cut back to smaller unpadded volume */ for( pp=kk=0 ; kk < nz ; kk++ ){ for( jj=0 ; jj < ny ; jj++ ){ qq = ex_pad + (jj+ey_pad)*nx_pad + (kk+ez_pad)*nxy_pad ; for( ii=0 ; ii < nx ; ii++ ) fim[pp++] = afim[qq++] ; }} /* free and clear whichever image we consumed here */ if( do_a ){ mri_free(acf_aim[ithr]) ; acf_aim[ithr] = NULL ; } else { mri_free(acf_bim[ithr]) ; acf_bim[ithr] = NULL ; } return ; } /*---------------------------------------------------------------------------*/ /* Create the "functional" image, with smoothing and padding [12 May 2015] */ void generate_fim_padded( float *fim , float *pfim , unsigned short xran[] ) { int ii,jj,kk,pp,qq ; /* random N(0,1) stuff to create the larger (padded) volume */ for( ii=0 ; ii < nxyz_pad ; ii++ ) pfim[ii] = zgaussian_sss(xran) ; /* smoothization */ if( do_blur ) FIR_blur_volume_3d(nx_pad,ny_pad,nz_pad,dx,dy,dz,pfim,sigmax,sigmay,sigmaz) ; /* cut back to smaller unpadded volume */ for( pp=kk=0 ; kk < nz ; kk++ ){ for( jj=0 ; jj < ny ; jj++ ){ qq = ex_pad + (jj+ey_pad)*nx_pad + (kk+ez_pad)*nxy_pad ; for( ii=0 ; ii < nx ; ii++ ) fim[pp++] = pfim[qq++] ; }} return ; } /*---------------------------------------------------------------------------*/ /* Create the functional "image", with no padding and optional smoothing. */ void generate_fim_unpadded( float *fim , unsigned short xran[] ) { int ii ; float sum ; /* random N(0,1) stuff */ for( ii=0 ; ii < nxyz ; ii++ ) fim[ii] = zgaussian_sss(xran) ; /* smoothization */ if( do_blur ) FIR_blur_volume_3d(nx,ny,nz,dx,dy,dz,fim,sigmax,sigmay,sigmaz) ; return ; } /*---------------------------------------------------------------------------*/ /* Generate random smoothed masked image, with stdev=1. */ void generate_image( float *fim , float *pfim , unsigned short xran[] , int iter ) { register int ii ; register float sum ; /* Outsource the creation of the smoothed random field [12 May 2015] */ if( inset != NULL ) generate_fim_inset( fim , iter-1 ) ; else if( xinset != NULL ) generate_fim_insdat( fim , iter-1 ) ; else if( do_acf ) generate_fim_acf( fim , xran ) ; else if( do_pad ) generate_fim_padded( fim , pfim , xran ) ; else generate_fim_unpadded( fim , xran ) ; /* normalizing */ if( num_inset == 0 && xinset == NULL ){ for( sum=0.0f,ii=0 ; ii < nxyz ; ii++ ) sum += fim[ii]*fim[ii] ; sum = sqrtf( nxyz / sum ) ; /* fix stdev back to 1 */ for( ii=0 ; ii < nxyz ; ii++ ) fim[ii] *= sum ; } /* save this volume? */ if( do_ssave == SSAVE_BLURRED ) ssave_dataset(fim) ; /* maskizing? */ if( mask_vol != NULL && xinset == NULL ){ for( ii=0 ; ii < nxyz ; ii++ ) if( !mask_vol[ii] ) fim[ii] = 0.0f ; } if( num_inset == 0 && xinset == NULL && tdof > 0.0f ){ float zfac = 1.0f/(1.0f-0.25f/tdof) ; /* 26 May 2015: secret stuff */ float tfac = 0.5f/tdof ; float zhat , denom ; for( ii=0 ; ii < nxyz ; ii++ ){ if( fim[ii] != 0.0f ){ zhat = fim[ii]*zfac ; denom = 1.0f - zhat*zhat*tfac ; if( denom < 0.1f ) denom = 0.1f ; fim[ii] = zhat / sqrtf(denom) ; } } } /* save THIS volume? */ if( do_ssave == SSAVE_MASKED ) ssave_dataset(fim) ; #if 0 /* absolution? */ if( do_2sid ){ for( ii=0 ; ii < nxyz ; ii++ ) fim[ii] = fabsf(fim[ii]) ; } #endif return ; } #ifdef USE_SHAVE /*---------------------------------------------------------------------------*/ /* 'shave' == 'short save' == saving generated images for reuse in the sumup phase of the program. Only the "in-mask" part is saved, and values are scale to shorts. Memory is malloc-ed for small needs, and is mmap-ed for large needs. *//*-------------------------------------------------------------------------*/ #include #if !defined(MAP_ANON) && defined(MAP_ANONYMOUS) # define MAP_ANON MAP_ANONYMOUS #endif void setup_shave(void) { int64_t twogig = 2ll * 1024ll * 1024ll * 1024ll ; ENTRY("setup_shave") ; shave_siz = (size_t)mask_ngood ; /* in units of sizeof(short) */ shave_tot = shave_siz * (int64_t)niter * sizeof(short) ; /* check is system is 32-bit and memory needed is over 2G */ if( shave_tot >= twogig && ( sizeof(void *) < 8 || sizeof(size_t) < 8 ) ) ERROR_exit("Total space needed for internal save of simulations\n" " exceeds 2 GB -- cannot proceed on a 32-bit system!") ; #ifndef MAP_ANON do_shave = SHAVE_MALLOC #else do_shave = (shave_tot >= twogig) ? SHAVE_MMAP : SHAVE_MALLOC ; #endif if( do_shave == SHAVE_MALLOC ){ shave = (short *)malloc((size_t)shave_tot) ; } else { #ifdef MAP_ANON shave = mmap( (void *)0 , (size_t)shave_tot , PROT_READ | PROT_WRITE , MAP_ANON | MAP_SHARED , -1,0 ) ; #endif } if( shave == NULL ) ERROR_exit("Cannot allocate space for internal save of simulations :-(") ; if( verb ) INFO_message("allocated %s (%s) bytes for sumup reuse", commaized_integer_string((long long)shave_tot) , approximate_number_string((double)shave_tot) ) ; EXRETURN ; } /*---------------------------------------------------------------------------*/ /* Trash the allocated shave space */ void destroy_shave(void) { ENTRY("destroy_shave") ; if( shave == NULL ) EXRETURN ; if( do_shave == SHAVE_MALLOC ) free(shave) ; else munmap(shave,(size_t)shave_tot) ; shave = NULL ; EXRETURN ; } /*---------------------------------------------------------------------------*/ /* Actually save the iter-th image into the allocated space */ #define SHAVE_FAC 4681.0f /* 32767/7 */ #define SHAVE_INV 0.0002136296f /* 7/32767 */ void fim_to_shave( float *fim , int iter ) { short *shar = shave + (size_t)(iter * shave_siz) ; int ii , jj ; #if 0 #pragma omp critical { fprintf(stderr,"f2s(%d) ",iter); } #endif if( mask_vol != NULL ){ for( jj=ii=0 ; ii < nxyz ; ii++ ) if( mask_vol[ii] ) shar[jj++] = (short)(fim[ii]*SHAVE_FAC) ; } else { for( ii=0 ; ii < nxyz ; ii++ ) shar[ii] = (short)(fim[ii]*SHAVE_FAC) ; } return ; } /*---------------------------------------------------------------------------*/ /* Retrieve the iter-th image from the allocated space */ void shave_to_fim( float *fim , int iter ) { short *shar = shave + (size_t)(iter * shave_siz) ; int ii , jj ; #if 0 #pragma omp critical { fprintf(stderr,"s2f(%d) ",iter); } #endif if( mask_vol != NULL ){ for( jj=ii=0 ; ii < nxyz ; ii++ ){ if( mask_vol[ii] ) fim[ii] = shar[jj++]*SHAVE_INV ; else fim[ii] = 0.0f ; } } else { for( ii=0 ; ii < nxyz ; ii++ ) fim[ii] = shar[ii]*SHAVE_INV ; } return ; } #endif /*---------------------------------------------------------------------------*/ #define DALL MAX_CLUSTER_SIZE /*! Put (i,j,k) into the current cluster, if it is nonzero. */ #define CPUT(i,j,k) \ do{ ijk = THREE_TO_IJK(i,j,k,nx,nxy) ; \ if( mmm[ijk] ){ \ if( nnow == nall ){ /* increase array lengths */ \ nall += DALL + nall/2 ; \ inow = (short *) realloc(inow,sizeof(short)*nall) ; \ jnow = (short *) realloc(jnow,sizeof(short)*nall) ; \ know = (short *) realloc(know,sizeof(short)*nall) ; \ } \ inow[nnow] = i ; jnow[nnow] = j ; know[nnow] = k ; \ nnow++ ; mmm[ijk] = 0 ; \ } } while(0) #define USE_MEMCHR static int *nall_g = NULL ; /* per-thread workspaces */ static short **inow_g = NULL ; /* for clusterizationing */ static short **jnow_g = NULL ; static short **know_g = NULL ; /*----------------------------------------------------------------------------*/ int find_largest_cluster_NN1( byte *mmm , int ithr ) { int max_size=0 ; int ii,jj,kk, icl , ijk , ijk_last ; int ip,jp,kp , im,jm,km ; int nnow , nall ; short *inow , *jnow , *know ; #ifdef USE_MEMCHR byte *mch ; #endif /* get workspace for this thread */ nall = nall_g[ithr] ; inow = inow_g[ithr] ; jnow = jnow_g[ithr] ; know = know_g[ithr] ; ijk_last = 0 ; /* start scanning at the start */ while(1) { /* find next nonzero point in mmm array */ #ifndef USE_MEMCHR for( ijk=ijk_last ; ijk < nxyz ; ijk++ ) if( mmm[ijk] ) break ; #else mch = memchr( mmm+ijk_last , 1 , nxyz-ijk_last ) ; /* quicker search */ if( mch == NULL ) ijk = nxyz ; else ijk = mch - mmm ; #endif if( ijk == nxyz ) break ; /* didn't find any! */ ijk_last = ijk+1 ; /* start here next time */ mmm[ijk] = 0 ; /* clear found point */ nnow = 1 ; /* # pts in cluster */ IJK_TO_THREE(ijk, inow[0],jnow[0],know[0] , nx,nxy) ; /* loop over points in cluster, checking their neighbors, growing the cluster if we find any that belong therein */ for( icl=0 ; icl < nnow ; icl++ ){ ii = inow[icl] ; jj = jnow[icl] ; kk = know[icl] ; im = ii-1 ; jm = jj-1 ; km = kk-1 ; ip = ii+1 ; jp = jj+1 ; kp = kk+1 ; if( im >= 0 ) CPUT(im,jj,kk) ; if( ip < nx ) CPUT(ip,jj,kk) ; if( jm >= 0 ) CPUT(ii,jm,kk) ; if( jp < ny ) CPUT(ii,jp,kk) ; if( km >= 0 ) CPUT(ii,jj,km) ; if( kp < nz ) CPUT(ii,jj,kp) ; } if( nnow > max_size ) max_size = nnow ; } if( nall > nall_g[ithr] ){ /* probably won't happen */ nall_g[ithr] = nall ; inow_g[ithr] = inow ; jnow_g[ithr] = jnow ; know_g[ithr] = know ; } return max_size ; } /*----------------------------------------------------------------------------*/ int find_largest_cluster_NN2( byte *mmm , int ithr ) { int max_size=0 ; int ii,jj,kk, icl , ijk , ijk_last ; int ip,jp,kp , im,jm,km ; int nnow , nall ; short *inow , *jnow , *know ; #ifdef USE_MEMCHR byte *mch ; #endif /* get workspace for this thread */ nall = nall_g[ithr] ; inow = inow_g[ithr] ; jnow = jnow_g[ithr] ; know = know_g[ithr] ; ijk_last = 0 ; /* start scanning at the start */ while(1) { /* find next nonzero point in mmm array */ #ifndef USE_MEMCHR for( ijk=ijk_last ; ijk < nxyz ; ijk++ ) if( mmm[ijk] ) break ; #else mch = memchr( mmm+ijk_last , 1 , nxyz-ijk_last ) ; /* quicker search */ if( mch == NULL ) ijk = nxyz ; else ijk = mch - mmm ; #endif if( ijk == nxyz ) break ; /* didn't find any! */ ijk_last = ijk+1 ; /* start here next time */ mmm[ijk] = 0 ; /* clear found point */ nnow = 1 ; /* # pts in cluster */ IJK_TO_THREE(ijk, inow[0],jnow[0],know[0] , nx,nxy) ; /* loop over points in cluster, checking their neighbors, growing the cluster if we find any that belong therein */ for( icl=0 ; icl < nnow ; icl++ ){ ii = inow[icl] ; jj = jnow[icl] ; kk = know[icl] ; im = ii-1 ; jm = jj-1 ; km = kk-1 ; ip = ii+1 ; jp = jj+1 ; kp = kk+1 ; if( im >= 0 ){ CPUT(im,jj,kk) ; if( jm >= 0 ) CPUT(im,jm,kk) ; /* 2NN */ if( jp < ny ) CPUT(im,jp,kk) ; /* 2NN */ if( km >= 0 ) CPUT(im,jj,km) ; /* 2NN */ if( kp < nz ) CPUT(im,jj,kp) ; /* 2NN */ } if( ip < nx ){ CPUT(ip,jj,kk) ; if( jm >= 0 ) CPUT(ip,jm,kk) ; /* 2NN */ if( jp < ny ) CPUT(ip,jp,kk) ; /* 2NN */ if( km >= 0 ) CPUT(ip,jj,km) ; /* 2NN */ if( kp < nz ) CPUT(ip,jj,kp) ; /* 2NN */ } if( jm >= 0 ){ CPUT(ii,jm,kk) ; if( km >= 0 ) CPUT(ii,jm,km) ; /* 2NN */ if( kp < nz ) CPUT(ii,jm,kp) ; /* 2NN */ } if( jp < ny ){ CPUT(ii,jp,kk) ; if( km >= 0 ) CPUT(ii,jp,km) ; /* 2NN */ if( kp < nz ) CPUT(ii,jp,kp) ; /* 2NN */ } if( km >= 0 ) CPUT(ii,jj,km) ; if( kp < nz ) CPUT(ii,jj,kp) ; } if( nnow > max_size ) max_size = nnow ; } if( nall > nall_g[ithr] ){ /* probably won't happen */ nall_g[ithr] = nall ; inow_g[ithr] = inow ; jnow_g[ithr] = jnow ; know_g[ithr] = know ; } return max_size ; } /*----------------------------------------------------------------------------*/ int find_largest_cluster_NN3( byte *mmm , int ithr ) { int max_size=0 ; int ii,jj,kk, icl , ijk , ijk_last ; int ip,jp,kp , im,jm,km ; int nnow , nall ; short *inow , *jnow , *know ; #ifdef USE_MEMCHR byte *mch ; #endif /* get workspace for this thread */ nall = nall_g[ithr] ; inow = inow_g[ithr] ; jnow = jnow_g[ithr] ; know = know_g[ithr] ; ijk_last = 0 ; /* start scanning at the start */ while(1) { /* find next nonzero point in mmm array */ #ifndef USE_MEMCHR for( ijk=ijk_last ; ijk < nxyz ; ijk++ ) if( mmm[ijk] ) break ; #else mch = memchr( mmm+ijk_last , 1 , nxyz-ijk_last ) ; /* quicker search */ if( mch == NULL ) ijk = nxyz ; else ijk = mch - mmm ; #endif if( ijk == nxyz ) break ; /* didn't find any! */ ijk_last = ijk+1 ; /* start here next time */ mmm[ijk] = 0 ; /* clear found point */ nnow = 1 ; /* # pts in cluster */ IJK_TO_THREE(ijk, inow[0],jnow[0],know[0] , nx,nxy) ; /* loop over points in cluster, checking their neighbors, growing the cluster if we find any that belong therein */ for( icl=0 ; icl < nnow ; icl++ ){ ii = inow[icl] ; jj = jnow[icl] ; kk = know[icl] ; im = ii-1 ; jm = jj-1 ; km = kk-1 ; ip = ii+1 ; jp = jj+1 ; kp = kk+1 ; if( im >= 0 ){ CPUT(im,jj,kk) ; if( jm >= 0 ) CPUT(im,jm,kk) ; /* 2NN */ if( jp < ny ) CPUT(im,jp,kk) ; /* 2NN */ if( km >= 0 ) CPUT(im,jj,km) ; /* 2NN */ if( kp < nz ) CPUT(im,jj,kp) ; /* 2NN */ if( jm >= 0 && km >= 0 ) CPUT(im,jm,km) ; /* 3NN */ if( jm >= 0 && kp < nz ) CPUT(im,jm,kp) ; /* 3NN */ if( jp < ny && km >= 0 ) CPUT(im,jp,km) ; /* 3NN */ if( jp < ny && kp < nz ) CPUT(im,jp,kp) ; /* 3NN */ } if( ip < nx ){ CPUT(ip,jj,kk) ; if( jm >= 0 ) CPUT(ip,jm,kk) ; /* 2NN */ if( jp < ny ) CPUT(ip,jp,kk) ; /* 2NN */ if( km >= 0 ) CPUT(ip,jj,km) ; /* 2NN */ if( kp < nz ) CPUT(ip,jj,kp) ; /* 2NN */ if( jm >= 0 && km >= 0 ) CPUT(ip,jm,km) ; /* 3NN */ if( jm >= 0 && kp < nz ) CPUT(ip,jm,kp) ; /* 3NN */ if( jp < ny && km >= 0 ) CPUT(ip,jp,km) ; /* 3NN */ if( jp < ny && kp < nz ) CPUT(ip,jp,kp) ; /* 3NN */ } if( jm >= 0 ){ CPUT(ii,jm,kk) ; if( km >= 0 ) CPUT(ii,jm,km) ; /* 2NN */ if( kp < nz ) CPUT(ii,jm,kp) ; /* 2NN */ } if( jp < ny ){ CPUT(ii,jp,kk) ; if( km >= 0 ) CPUT(ii,jp,km) ; /* 2NN */ if( kp < nz ) CPUT(ii,jp,kp) ; /* 2NN */ } if( km >= 0 ) CPUT(ii,jj,km) ; if( kp < nz ) CPUT(ii,jj,kp) ; } if( nnow > max_size ) max_size = nnow ; } if( nall > nall_g[ithr] ){ /* probably won't happen */ nall_g[ithr] = nall ; inow_g[ithr] = inow ; jnow_g[ithr] = jnow ; know_g[ithr] = know ; } return max_size ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN1_1sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN1( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_1sid_NN1[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN2_1sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN2( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_1sid_NN2[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN3_1sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN3( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_1sid_NN3[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN1_2sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN1( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_2sid_NN1[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN2_2sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN2( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_2sid_NN2[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN3_2sid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN3( bfim , ithr ) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_2sid_NN3[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN1_bsid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz_p , siz_m , siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz_p = find_largest_cluster_NN1( bfim , ithr ) ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz_m = find_largest_cluster_NN1( bfim , ithr ) ; siz = MAX(siz_p,siz_m) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_bsid_NN1[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN2_bsid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz_p , siz_m , siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz_p = find_largest_cluster_NN2( bfim , ithr ) ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz_m = find_largest_cluster_NN2( bfim , ithr ) ; siz = MAX(siz_p,siz_m) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_bsid_NN2[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ /* Find clusters, save some info */ void gather_stats_NN3_bsid( int ipthr, float *fim, byte *bfim, int *mtab, int ithr,int iter ) { register int ii ; register float thr ; int siz_p , siz_m , siz ; thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz_p = find_largest_cluster_NN3( bfim , ithr ) ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz_m = find_largest_cluster_NN3( bfim , ithr ) ; siz = MAX(siz_p,siz_m) ; if( siz > max_cluster_size ) siz = max_cluster_size ; mtab[siz]++ ; csiz_bsid_NN3[ipthr][iter-1] = siz ; return ; } /*---------------------------------------------------------------------------*/ static char * prob6(float p) /* format p-value into 6 char */ { static char str[16] ; if( p >= 0.00010f ){ sprintf(str,"%7.5f",p) ; } else { int dec = (int)(0.9999-log10(p)) ; float man = p * pow( 10.0 , (double)dec ) ; sprintf(str,"%4.1fe-%1d",man,dec) ; } return (str+1) ; } /*---------------------------------------------------------------------------*/ static char * prob9(float p) /* format p-value into 9 char */ { static char str[16] ; if( p >= 0.00010f ){ sprintf(str,"%9.6f",p) ; } else { int dec = (int)(0.9999-log10(p)) ; float man = p * pow( 10.0 , (double)dec ) ; sprintf(str,"%6.3fe-%1d",man,dec) ; } return str ; } /*---------------------------------------------------------------------------*/ #ifdef USE_SHAVE static int *fa_1sid_NN1, *fa_1sid_NN2, *fa_1sid_NN3 ; static int *fa_2sid_NN1, *fa_2sid_NN2, *fa_2sid_NN3 ; static int *fa_bsid_NN1, *fa_bsid_NN2, *fa_bsid_NN3 ; void thresh_summer_fim( int iathr, int ipthr_bot, int ipthr_top, float *fim, byte *bfim, int ithr ) { register int ii ; register float thr ; int ipthr , siz ; fa_1sid_NN1[ithr]=0; fa_1sid_NN2[ithr]=0; fa_1sid_NN3[ithr]=0; fa_2sid_NN1[ithr]=0; fa_2sid_NN2[ithr]=0; fa_2sid_NN3[ithr]=0; fa_bsid_NN1[ithr]=0; fa_bsid_NN2[ithr]=0; fa_bsid_NN3[ithr]=0; for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ if( !fa_1sid_NN1[ithr] ){ /* 1-sided, NN1 */ thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN1(bfim,0) ; fa_1sid_NN1[ithr] = (siz >= clust_thresh_1sid_NN1[ipthr][iathr] ) ; } if( !fa_1sid_NN2[ithr] ){ /* 1-sided, NN2 */ thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN2(bfim,0) ; fa_1sid_NN2[ithr] = (siz >= clust_thresh_1sid_NN2[ipthr][iathr] ) ; } if( !fa_1sid_NN3[ithr] ){ /* 1-sided, NN3 */ thr = zthr_1sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN3(bfim,0) ; fa_1sid_NN3[ithr] = (siz >= clust_thresh_1sid_NN3[ipthr][iathr] ) ; } if( do_athr_sum < 2 ) continue ; /* skip 2-sided and bi-sided */ if( !fa_2sid_NN1[ithr] ){ /* 2-sided, NN1 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN1(bfim,0) ; fa_2sid_NN1[ithr] = (siz >= clust_thresh_2sid_NN1[ipthr][iathr] ) ; } if( !fa_2sid_NN2[ithr] ){ /* 2-sided, NN2 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN2(bfim,0) ; fa_2sid_NN2[ithr] = (siz >= clust_thresh_2sid_NN2[ipthr][iathr] ) ; } if( !fa_2sid_NN3[ithr] ){ /* 2-sided, NN3 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) || (fim[ii] < -thr) ; siz = find_largest_cluster_NN3(bfim,0) ; fa_2sid_NN3[ithr] = (siz >= clust_thresh_2sid_NN3[ipthr][iathr] ) ; } if( !fa_bsid_NN1[ithr] ){ /* bi-sided, NN1 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN1( bfim , 0 ) ; fa_bsid_NN1[ithr] = (siz >= clust_thresh_bsid_NN1[ipthr][iathr] ) ; if( !fa_bsid_NN1[ithr] ){ for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz = find_largest_cluster_NN1( bfim , 0 ) ; fa_bsid_NN1[ithr] = (siz >= clust_thresh_bsid_NN1[ipthr][iathr] ) ; } } if( !fa_bsid_NN2[ithr] ){ /* bi-sided, NN2 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN2( bfim , 0 ) ; fa_bsid_NN2[ithr] = (siz >= clust_thresh_bsid_NN2[ipthr][iathr] ) ; if( !fa_bsid_NN2[ithr] ){ for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz = find_largest_cluster_NN2( bfim , 0 ) ; fa_bsid_NN2[ithr] = (siz >= clust_thresh_bsid_NN2[ipthr][iathr] ) ; } } if( !fa_bsid_NN3[ithr] ){ /* bi-sided, NN3 */ thr = zthr_2sid[ipthr] ; for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] > thr) ; siz = find_largest_cluster_NN3( bfim , 0 ) ; fa_bsid_NN3[ithr] = (siz >= clust_thresh_bsid_NN3[ipthr][iathr] ) ; if( !fa_bsid_NN3[ithr] ){ for( ii=0 ; ii < nxyz ; ii++ ) bfim[ii] = (fim[ii] < -thr) ; siz = find_largest_cluster_NN3( bfim , 0 ) ; fa_bsid_NN3[ithr] = (siz >= clust_thresh_bsid_NN3[ipthr][iathr] ) ; } } } return ; } /*---------------------------------------------------------------------------*/ static float *rfa_1sid_NN1, *rfa_1sid_NN2, *rfa_1sid_NN3 ; static float *rfa_2sid_NN1, *rfa_2sid_NN2, *rfa_2sid_NN3 ; static float *rfa_bsid_NN1, *rfa_bsid_NN2, *rfa_bsid_NN3 ; void thresh_summer_athr( int ipthr_bot , int ipthr_top ) { float const drfa = 1.0f/niter ; ENTRY("thresh_summer_athr") ; rfa_1sid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_1sid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_1sid_NN3 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN3 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN3 = (float *)calloc(sizeof(float),nathr) ; AFNI_OMP_START; #pragma omp parallel { DECLARE_ithr ; int iter , iathr , vstep , vii ; float *fim ; byte *bfim ; #pragma omp master { #ifdef USE_OMP nthr = omp_get_num_threads() ; #else nthr = 1 ; #endif fa_1sid_NN1 = (int *)malloc(sizeof(int)*nthr) ; fa_2sid_NN1 = (int *)malloc(sizeof(int)*nthr) ; fa_bsid_NN1 = (int *)malloc(sizeof(int)*nthr) ; fa_1sid_NN2 = (int *)malloc(sizeof(int)*nthr) ; fa_2sid_NN2 = (int *)malloc(sizeof(int)*nthr) ; fa_bsid_NN2 = (int *)malloc(sizeof(int)*nthr) ; fa_1sid_NN3 = (int *)malloc(sizeof(int)*nthr) ; fa_2sid_NN3 = (int *)malloc(sizeof(int)*nthr) ; fa_bsid_NN3 = (int *)malloc(sizeof(int)*nthr) ; if( verb ){ vstep = (int)( niter / (nthr*50.0f) + 0.901f) ; vii = 0 ; vstep_reset() ; fprintf(stderr,"Summing up:") ; } } #pragma omp barrier fim = (float *)malloc(sizeof(float)*nxyz) ; bfim = (byte *)malloc(sizeof(byte) *nxyz) ; #pragma omp for for( iter=1 ; iter <= niter ; iter++ ){ if( ithr==0 && verb ){ vii++ ; if( vii%vstep == vstep/2 ) vstep_print() ; } shave_to_fim(fim,iter-1) ; for( iathr=0 ; iathr < nathr ; iathr++ ){ thresh_summer_fim( iathr,ipthr_bot,fim,bfim,ithr ) ; #pragma omp critical {if( fa_1sid_NN1[ithr] ) rfa_1sid_NN1[iathr] += drfa ; if( fa_1sid_NN2[ithr] ) rfa_1sid_NN2[iathr] += drfa ; if( fa_1sid_NN3[ithr] ) rfa_1sid_NN3[iathr] += drfa ; if( do_athr_sum > 1 ){ if( fa_2sid_NN1[ithr] ) rfa_2sid_NN1[iathr] += drfa ; if( fa_2sid_NN2[ithr] ) rfa_2sid_NN2[iathr] += drfa ; if( fa_2sid_NN3[ithr] ) rfa_2sid_NN3[iathr] += drfa ; if( fa_bsid_NN1[ithr] ) rfa_bsid_NN1[iathr] += drfa ; if( fa_bsid_NN2[ithr] ) rfa_bsid_NN2[iathr] += drfa ; if( fa_bsid_NN3[ithr] ) rfa_bsid_NN3[iathr] += drfa ; } } } } free(bfim) ; free(fim) ; } AFNI_OMP_END ; if( verb ) fprintf(stderr,"\n") ; EXRETURN ; } #else /*---------------------------------------------------------------------------*/ static float *rfa_1sid_NN1, *rfa_1sid_NN2, *rfa_1sid_NN3 ; static float *rfa_2sid_NN1, *rfa_2sid_NN2, *rfa_2sid_NN3 ; static float *rfa_bsid_NN1, *rfa_bsid_NN2, *rfa_bsid_NN3 ; void thresh_summer_onecurve( int iathr , int ipthr_bot , int ipthr_top ) { float const drfa = 1.0f/niter ; int ipthr, siz , iter ; ENTRY("thresh_summer_onecurve") ; for( iter=0 ; iter < niter ; iter++ ){ for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_1sid_NN1[ipthr][iter] ; if( siz >= clust_thresh_1sid_NN1[ipthr][iathr] ){ rfa_1sid_NN1[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_1sid_NN2[ipthr][iter] ; if( siz >= clust_thresh_1sid_NN2[ipthr][iathr] ){ rfa_1sid_NN2[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_1sid_NN3[ipthr][iter] ; if( siz >= clust_thresh_1sid_NN3[ipthr][iathr] ){ rfa_1sid_NN3[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_2sid_NN1[ipthr][iter] ; if( siz >= clust_thresh_2sid_NN1[ipthr][iathr] ){ rfa_2sid_NN1[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_2sid_NN2[ipthr][iter] ; if( siz >= clust_thresh_2sid_NN2[ipthr][iathr] ){ rfa_2sid_NN2[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_2sid_NN3[ipthr][iter] ; if( siz >= clust_thresh_2sid_NN3[ipthr][iathr] ){ rfa_2sid_NN3[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_bsid_NN1[ipthr][iter] ; if( siz >= clust_thresh_bsid_NN1[ipthr][iathr] ){ rfa_bsid_NN1[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_bsid_NN2[ipthr][iter] ; if( siz >= clust_thresh_bsid_NN2[ipthr][iathr] ){ rfa_bsid_NN2[iathr] += drfa ; break ; } } for( ipthr=ipthr_bot ; ipthr <= ipthr_top ; ipthr++ ){ siz = csiz_bsid_NN3[ipthr][iter] ; if( siz >= clust_thresh_bsid_NN3[ipthr][iathr] ){ rfa_bsid_NN3[iathr] += drfa ; break ; } } } EXRETURN ; } /*---------------------------------------------------------------------------*/ void thresh_summer_athr( int ipthr_bot , int ipthr_top ) { int iathr ; if( ipthr_top < ipthr_bot ) ipthr_top = npthr-1 ; ENTRY("thresh_summer_athr") ; rfa_1sid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN1 = (float *)calloc(sizeof(float),nathr) ; rfa_1sid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN2 = (float *)calloc(sizeof(float),nathr) ; rfa_1sid_NN3 = (float *)calloc(sizeof(float),nathr) ; rfa_2sid_NN3 = (float *)calloc(sizeof(float),nathr) ; rfa_bsid_NN3 = (float *)calloc(sizeof(float),nathr) ; for( iathr=0 ; iathr < nathr ; iathr++ ){ thresh_summer_onecurve(iathr,ipthr_bot,ipthr_top) ; } EXRETURN ; } #endif /*---------------------------------------------------------------------------*/ /* from the max_table, get the threshold cluster size for a given per-voxel threshold, for a given false alarm rate (alpha). *//*-------------------------------------------------------------------------*/ double get_one_clust_thresh( int **mtnn , int ipthr , double aval ) { static double *alpha=NULL ; double ahi,alo,jj ; float cmax=0.0f , **clust_thresh ; int ii , itop , iathr , mmm ; if( alpha == NULL ) alpha = (double *)malloc(sizeof(double)*(max_cluster_size+1)) ; /* alpha[ii] = prob of getting clusters of exactly size ii */ for( itop=ii=1 ; ii <= max_cluster_size ; ii++ ){ alpha[ii] = mtnn[ipthr][ii] / (double)niter ; if( alpha[ii] > 0.0 ) itop = ii ; } /* alpha[ii] = prob of getting clusters of size ii or more */ for( ii=itop-1 ; ii >= 1 ; ii-- ) alpha[ii] += alpha[ii+1] ; /* find ii that brackets the desired aval */ if( aval > alpha[1] ){ /* unpleasant situation */ static int first=2 ; /* where there is no bracket */ ii = 1 ; if( first ){ WARNING_message( "pthr=%.6f ; desired alpha=%.6f -- but max simulated alpha=%.6f]\n" , pthr[ipthr] , aval , alpha[1] ) ; first-- ; if( first == 0 ) ININFO_message(" [further messages about alpha are suppressed]") ; } } else { for( ii=1 ; ii < itop ; ii++ ){ if( alpha[ii] >= aval && alpha[ii+1] <= aval ) break ; } } /* inverse interpolate to find the index jj where alpha[jj]=aval -- except, of course, jj is a floating point value between ii and ii+1 */ alo=alpha[ii] ; ahi=alpha[ii+1] ; if( alo >= 1.0 ) alo = 1.0 - 0.1/niter ; /* unlikely */ if( ahi <= 0.0 ) ahi = 0.1/niter ; /* should not be possible */ if( ahi >= alo ) ahi = 0.1*alo ; /* should not be possible */ jj = log(-log(1.0-aval)) ; alo = log(-log(1.0-alo)) ; ahi = log(-log(1.0-ahi)) ; jj = ii + (jj-alo)/(ahi-alo) ; if( jj < 1.0 ) jj = 1.0 ; return jj ; } /*===========================================================================*/ int main( int argc , char **argv ) { int nnn , ipthr , first_mask=1 ; char *refit_cmd = NULL ; double ct ; #ifdef USE_OMP int ***mtab_1sid[4] , ***mtab_2sid[4] , ***mtab_bsid[4] ; #endif /*----- does user request help menu? -----*/ AFNI_SETUP_OMP(0) ; /* 24 Jun 2013 */ (void)COX_clock_time() ; if( argc < 2 || strcmp(argv[1],"-help") == 0 ) display_help_menu() ; /*----- get the list of things to do -----*/ mainENTRY("3dClustSim"); machdep(); AFNI_logger("3dClustSim",argc,argv); PRINT_VERSION("3dClustSim"); AUTHOR("RW Cox and BD Ward"); THD_check_AFNI_version("3dClustSim") ; get_options( argc , argv ) ; /*----- create some space for the results -----*/ for( nnn=1 ; nnn <= 3 ; nnn++ ){ max_table_1sid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; /* array of tables */ max_table_2sid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; /* array of tables */ max_table_bsid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; /* array of tables */ cmx_table_1sid[nnn] = (int * )malloc(sizeof(int )*npthr) ; /* 23 Dec 2015 */ cmx_table_2sid[nnn] = (int * )malloc(sizeof(int )*npthr) ; cmx_table_bsid[nnn] = (int * )malloc(sizeof(int )*npthr) ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ /* create tables */ max_table_1sid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; max_table_2sid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; max_table_bsid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; } } csiz_1sid_NN1 = (int **)malloc(sizeof(int *)*npthr) ; /* 23 Dec 2015 */ csiz_1sid_NN2 = (int **)malloc(sizeof(int *)*npthr) ; csiz_1sid_NN3 = (int **)malloc(sizeof(int *)*npthr) ; csiz_2sid_NN1 = (int **)malloc(sizeof(int *)*npthr) ; csiz_2sid_NN2 = (int **)malloc(sizeof(int *)*npthr) ; csiz_2sid_NN3 = (int **)malloc(sizeof(int *)*npthr) ; csiz_bsid_NN1 = (int **)malloc(sizeof(int *)*npthr) ; csiz_bsid_NN2 = (int **)malloc(sizeof(int *)*npthr) ; csiz_bsid_NN3 = (int **)malloc(sizeof(int *)*npthr) ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ csiz_1sid_NN1[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_1sid_NN2[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_1sid_NN3[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_2sid_NN1[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_2sid_NN2[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_2sid_NN3[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_bsid_NN1[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_bsid_NN2[ipthr] = (int *)calloc(sizeof(int),niter) ; csiz_bsid_NN3[ipthr] = (int *)calloc(sizeof(int),niter) ; } #ifdef USE_SHAVE if( do_athr_sum ) setup_shave() ; #endif if( verb ) INFO_message("Startup clock time = %.1f s",COX_clock_time()) ; AFNI_OMP_START ; #pragma omp parallel { DECLARE_ithr ; int iter, ipthr, **mt_1sid[4],**mt_2sid[4],**mt_bsid[4] , nnn ; float *fim ; byte *bfim ; unsigned short xran[3] ; float *pfim ; int vstep , vii ; /* create separate tables for each thread, if using OpenMP */ #ifdef USE_OMP #pragma omp master /* only in the master thread */ { nthr = omp_get_num_threads() ; for( nnn=1 ; nnn <= 3 ; nnn++ ){ mtab_1sid[nnn] = (int ***)malloc(sizeof(int **) *nthr) ; mtab_2sid[nnn] = (int ***)malloc(sizeof(int **) *nthr) ; mtab_bsid[nnn] = (int ***)malloc(sizeof(int **) *nthr) ; } nall_g = (int *) malloc(sizeof(int) *nthr) ; /* workspaces for */ inow_g = (short **)malloc(sizeof(short *)*nthr) ; /* find_largest_cluster() */ jnow_g = (short **)malloc(sizeof(short *)*nthr) ; know_g = (short **)malloc(sizeof(short *)*nthr) ; if( do_acf ){ acf_aim = (MRI_IMAGE **)calloc(sizeof(MRI_IMAGE *),nthr) ; acf_bim = (MRI_IMAGE **)calloc(sizeof(MRI_IMAGE *),nthr) ; acf_tar = (complex **) malloc(sizeof(complex *) *nthr) ; } if( verb ) INFO_message("Using %d OpenMP threads",nthr) ; } #pragma omp barrier /* all threads wait until the above is finished */ /* create tables for each thread separately */ for( nnn=1 ; nnn <= 3 ; nnn++ ){ mtab_1sid[nnn][ithr] = mt_1sid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; mtab_2sid[nnn][ithr] = mt_2sid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; mtab_bsid[nnn][ithr] = mt_bsid[nnn] = (int **)malloc(sizeof(int *)*npthr) ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ mt_1sid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; mt_2sid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; mt_bsid[nnn][ipthr] = (int *)calloc(sizeof(int),(max_cluster_size+1)) ; } } /* create workspace for find_largest_cluster(), for this thread */ nall_g[ithr] = DALL ; inow_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; jnow_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; know_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; if( do_acf ){ acf_tar[ithr] = (complex *)malloc(sizeof(complex)*acf_ntar) ; } /* initialize random seed array for each thread separately */ xran[2] = ( gseed & 0xffff) + (unsigned short)ithr ; xran[1] = ((gseed >> 16) & 0xffff) - (unsigned short)ithr ; xran[0] = 0x330e + (unsigned short)ithr ; #else /* not OpenMP ==> only one set of tables */ xran[2] = ( gseed & 0xffff) ; xran[1] = ((gseed >> 16) & 0xffff) ; xran[0] = 0x330e ; nall_g = (int *) malloc(sizeof(int) *nthr) ; inow_g = (short **)malloc(sizeof(short *)*nthr) ; jnow_g = (short **)malloc(sizeof(short *)*nthr) ; know_g = (short **)malloc(sizeof(short *)*nthr) ; nall_g[ithr] = DALL ; inow_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; jnow_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; know_g[ithr] = (short *) malloc(sizeof(short)*DALL) ; for( nnn=1 ; nnn <= 3 ; nnn++ ){ mt_1sid[nnn] = max_table_1sid[nnn] ; mt_2sid[nnn] = max_table_2sid[nnn] ; mt_bsid[nnn] = max_table_bsid[nnn] ; } if( do_acf ){ acf_aim = (MRI_IMAGE **)calloc(sizeof(MRI_IMAGE *),nthr) ; acf_bim = (MRI_IMAGE **)calloc(sizeof(MRI_IMAGE *),nthr) ; acf_tar = (complex **) calloc(sizeof(complex *) ,nthr) ; acf_tar[ithr] = (complex *)malloc(sizeof(complex)*acf_ntar) ; } #endif fim = (float *)malloc(sizeof(float)*nxyz) ; /* image space */ bfim = (byte * )malloc(sizeof(byte) *nxyz) ; if( do_pad ) pfim = (float *)malloc(sizeof(float)*nxyz_pad); /* 12 May 2015 */ else pfim = NULL ; vstep = (int)( niter / (nthr*50.0f) + 0.901f) ; vii = 0 ; if( ithr == 0 && verb ) fprintf(stderr,"Simulating:") ; /*----- Monte Carlo iterations -----*/ #pragma omp for for( iter=1 ; iter <= niter ; iter++ ){ if( verb && ithr == 0 ){ vii++ ; if( vii%vstep == 2 ) vstep_print() ; } generate_image( fim , pfim , xran , iter ) ; #ifdef USE_SHAVE if( do_shave ) fim_to_shave(fim,iter-1) ; #endif for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ gather_stats_NN1_1sid( ipthr, fim, bfim, mt_1sid[1][ipthr], ithr,iter ) ; gather_stats_NN2_1sid( ipthr, fim, bfim, mt_1sid[2][ipthr], ithr,iter ) ; gather_stats_NN3_1sid( ipthr, fim, bfim, mt_1sid[3][ipthr], ithr,iter ) ; gather_stats_NN1_2sid( ipthr, fim, bfim, mt_2sid[1][ipthr], ithr,iter ) ; gather_stats_NN2_2sid( ipthr, fim, bfim, mt_2sid[2][ipthr], ithr,iter ) ; gather_stats_NN3_2sid( ipthr, fim, bfim, mt_2sid[3][ipthr], ithr,iter ) ; gather_stats_NN1_bsid( ipthr, fim, bfim, mt_bsid[1][ipthr], ithr,iter ) ; gather_stats_NN2_bsid( ipthr, fim, bfim, mt_bsid[2][ipthr], ithr,iter ) ; gather_stats_NN3_bsid( ipthr, fim, bfim, mt_bsid[3][ipthr], ithr,iter ) ; } } /* end of simulation loop */ free(fim) ; free(bfim) ; if( pfim != NULL ) free(pfim) ; if( !do_athr_sum ){ free(inow_g[ithr]) ; free(jnow_g[ithr]) ; free(know_g[ithr]) ; if( do_acf ){ free(acf_tar[ithr]) ; if( acf_bim[ithr] != NULL ) mri_free(acf_bim[ithr]) ; } } if( ithr == 0 && verb ) fprintf(stderr,"!\n") ; } /* end OpenMP parallelization */ AFNI_OMP_END ; if( verb ) INFO_message("Clock time now = %.1f s",COX_clock_time()) ; if( do_acf && !do_athr_sum ){ free(acf_aim) ; free(acf_bim) ; free(acf_tar) ; } if( do_ssave > 0 ) DSET_delete(ssave_dset) ; /* 24 Apr 2014 */ /*-------- sum tables from various threads into one result ----------*/ #ifdef USE_OMP { int ithr , ii , ipthr , **mt_1sid,**mt_2sid,**mt_bsid , *mth_1sid,*mth_2sid,*mth_bsid ; for( nnn=1 ; nnn <= 3 ; nnn++ ){ for( ithr=0 ; ithr < nthr ; ithr++ ){ mt_1sid = mtab_1sid[nnn][ithr] ; mt_2sid = mtab_2sid[nnn][ithr] ; mt_bsid = mtab_bsid[nnn][ithr] ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ mth_1sid = mt_1sid[ipthr] ; mth_2sid = mt_2sid[ipthr] ; mth_bsid = mt_bsid[ipthr] ; for( ii=1 ; ii <= max_cluster_size ; ii++ ){ max_table_1sid[nnn][ipthr][ii] += mth_1sid[ii] ; max_table_2sid[nnn][ipthr][ii] += mth_2sid[ii] ; max_table_bsid[nnn][ipthr][ii] += mth_bsid[ii] ; } } } } } #endif /* For each table, find the largest cluster size cc such that max_table_xxxx[nnn][ipthr][cc] > 0 [23 Dec 2015] */ { int cc ; for( nnn=1 ; nnn <= 3 ; nnn++ ){ for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ for( cc=max_cluster_size ; cc >= 1 && max_table_1sid[nnn][ipthr][cc]==0 ; cc-- ) ; /*nada*/ cmx_table_1sid[nnn][ipthr] = cc ; for( cc=max_cluster_size ; cc >= 1 && max_table_2sid[nnn][ipthr][cc]==0 ; cc-- ) ; /*nada*/ cmx_table_2sid[nnn][ipthr] = cc ; for( cc=max_cluster_size ; cc >= 1 && max_table_bsid[nnn][ipthr][cc]==0 ; cc-- ) ; /*nada*/ cmx_table_bsid[nnn][ipthr] = cc ; } } } #if 1 if( AFNI_yesenv("AFNI_CLUSTSIM_SAVE") ){ FILE *fp; char fname[128]; static char *lnn[3] = { "NN1","NN2","NN3" }; int cc; for( nnn=1 ; nnn <= 3 ; nnn++ ){ for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ sprintf(fname,"mt.1sid.%s.%.5f.1D",lnn[nnn-1],pthr[ipthr]) ; fp = fopen(fname,"w") ; for( cc=1 ; cc <= cmx_table_1sid[nnn][ipthr] ; cc++ ) fprintf(fp,"%d %d\n",cc,max_table_1sid[nnn][ipthr][cc]) ; fclose(fp) ; sprintf(fname,"mt.2sid.%s.%.5f.1D",lnn[nnn-1],pthr[ipthr]) ; fp = fopen(fname,"w") ; for( cc=1 ; cc <= cmx_table_2sid[nnn][ipthr] ; cc++ ) fprintf(fp,"%d %d\n",cc,max_table_2sid[nnn][ipthr][cc]) ; fclose(fp) ; sprintf(fname,"mt.bsid.%s.%.5f.1D",lnn[nnn-1],pthr[ipthr]) ; fp = fopen(fname,"w") ; for( cc=1 ; cc <= cmx_table_bsid[nnn][ipthr] ; cc++ ) fprintf(fp,"%d %d\n",cc,max_table_bsid[nnn][ipthr][cc]) ; fclose(fp) ; } } } #endif #if 0 enable_mcw_malloc() ; #endif /*---------- compute and print the output tables ----------*/ { double *alpha , aval , ahi,alo ; float cmax=0.0f , **clust_thresh ; int ii , itop , iathr , mmm, ***mtt , **mtnn ; char *commandline = tross_commandline("3dClustSim",argc,argv) ; char fname[THD_MAX_NAME] , pname[THD_MAX_NAME] ; char *amesg = NULL , *mlab = NULL , *mlll = NULL ; alpha = (double *)malloc(sizeof(double)*(max_cluster_size+1)) ; for( mmm=1 ; mmm <= 3 ; mmm++ ){ /* loop over sidedness */ if( mmm == 1 ){ mtt = max_table_1sid ; mlab = "1-sided" ; mlll = "1sided" ; } else if( mmm == 2 ){ mtt = max_table_2sid ; mlab = "2-sided" ; mlll = "2sided" ; } else { mtt = max_table_bsid ; mlab = "bi-sided"; mlll = "bisided"; } amesg = NULL ; cmax = 0.0f ; for( nnn=1 ; nnn <= 3 ; nnn++ ){ /* loop over NN level */ clust_thresh = (float **)malloc(sizeof(float *)*npthr) ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ) clust_thresh[ipthr] = (float *)malloc(sizeof(float)*nathr) ; mtnn = mtt[nnn] ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ #if 1 for( iathr=0 ; iathr < nathr ; iathr++ ){ clust_thresh[ipthr][iathr] = get_one_clust_thresh( mtnn,ipthr,athr[iathr]) ; if( nodec ) clust_thresh[ipthr][iathr] = (int)(clust_thresh[ipthr][iathr]+0.951f) ; if( clust_thresh[ipthr][iathr] > cmax ) cmax = clust_thresh[ipthr][iathr] ; } #else for( itop=ii=1 ; ii <= max_cluster_size ; ii++ ){ alpha[ii] = mtt[nnn][ipthr][ii] / (double)niter ; if( alpha[ii] > 0.0 ) itop = ii ; } for( ii=itop-1 ; ii >= 1 ; ii-- ) alpha[ii] += alpha[ii+1] ; for( iathr=0 ; iathr < nathr ; iathr++ ){ aval = athr[iathr] ; if( aval > alpha[1] ){ /* unpleasant situation */ ii = 1 ; amesg = THD_zzprintf( amesg , " %s NN=%d pthr=%9.6f alpha=%6.3f [max simulated alpha=%6.3f]\n" , mlab , nnn , pthr[ipthr] , aval , alpha[1] ) ; } else { for( ii=1 ; ii < itop ; ii++ ){ if( alpha[ii] >= aval && alpha[ii+1] <= aval ) break ; } } alo=alpha[ii] ; ahi=alpha[ii+1] ; if( do_lohi ){ aval = ii ; /* for debugging */ } else { if( alo >= 1.0 ) alo = 1.0 - 0.1/niter ; if( ahi <= 0.0 ) ahi = 0.1/niter ; if( ahi >= alo ) ahi = 0.1*alo ; aval = log(-log(1.0-aval)) ; alo = log(-log(1.0-alo)) ; ahi = log(-log(1.0-ahi)) ; aval = ii + (aval-alo)/(ahi-alo) ; if( aval < 1.0 ) aval = 1.0 ; else if( nodec ) aval = (int)(aval+0.951) ; } clust_thresh[ipthr][iathr] = aval ; if( clust_thresh[ipthr][iathr] > cmax ) cmax = clust_thresh[ipthr][iathr] ; } #endif } /* end of loop over ipthr */ if( do_lohi == 0 ){ /* edit each column to increase as pthr increases [shouldn't be needed] */ for( iathr=0 ; iathr < nathr ; iathr++ ){ for( ipthr=npthr-2 ; ipthr >= 0 ; ipthr-- ){ if( clust_thresh[ipthr][iathr] < clust_thresh[ipthr+1][iathr] ) clust_thresh[ipthr][iathr] = clust_thresh[ipthr+1][iathr] ; } } /* edit each row to increase as athr decreases [shouldn't be needed] */ for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ for( iathr=1 ; iathr < nathr ; iathr++ ){ if( clust_thresh[ipthr][iathr] < clust_thresh[ipthr][iathr-1] ) clust_thresh[ipthr][iathr] = clust_thresh[ipthr][iathr-1] ; } } } #if 0 if( !nodec && !do_niml && cmax > 9999.9f ){ /* if largest is way big, */ for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ /* then truncate to ints. */ for( iathr=0 ; iathr < nathr ; iathr++ ){ aval = clust_thresh[ipthr][iathr] ; aval = (int)(aval+0.951) ; clust_thresh[ipthr][iathr] = aval ; } } nodec = 1 ; } #endif MPROBE ; if( prefix != NULL ){ sprintf(pname,"%s.NN%d_%s.",prefix,nnn,mlll) ; } else { fflush(stderr) ; fflush(stdout) ; } if( do_1D ){ /* output in 1D format */ FILE *fp = stdout ; if( prefix != NULL ){ strcpy(fname,pname) ; strcat(fname,"1D") ; fp = fopen(fname,"w") ; if( fp == NULL ){ ERROR_message("Can't open file %s -- using stdout",fname) ; fp = stdout ; } } fprintf(fp, "# %s\n" "# %s thresholding\n" "# Grid: %dx%dx%d %.2fx%.2fx%.2f mm^3 (%d voxels%s)\n" "#\n" "# CLUSTER SIZE THRESHOLD(pthr,alpha) in Voxels\n" "# -NN %d | alpha = Prob(Cluster >= given size)\n" "# pthr |" , commandline , mlab , nx,ny,nz , dx,dy,dz , mask_ngood , (mask_ngood < nxyz) ? " in mask" : "\0" , nnn ) ; for( iathr=0 ; iathr < nathr ; iathr++ ) fprintf(fp," %s",prob6(athr[iathr])) ; fprintf(fp,"\n" "# ------ |" ) ; for( iathr=0 ; iathr < nathr ; iathr++ ) fprintf(fp," ------") ; fprintf(fp,"\n") ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ){ fprintf(fp,"%s ",prob9(pthr[ipthr])) ; for( iathr=0 ; iathr < nathr ; iathr++ ){ if( nodec ) fprintf(fp,"%7d" ,(int)clust_thresh[ipthr][iathr]) ; else if( clust_thresh[ipthr][iathr] <= 9999.9f ) fprintf(fp,"%7.1f", clust_thresh[ipthr][iathr]) ; else fprintf(fp,"%7.0f", clust_thresh[ipthr][iathr]) ; } fprintf(fp,"\n") ; } if( fp != stdout ) fclose(fp) ; } if( do_niml ){ /* output in NIML format */ NI_element *nel ; float *vec ; char buf[1024] , *bbb ; NI_float_array nfar ; sprintf(buf,"3dClustSim_NN%d",nnn) ; nel = NI_new_data_element( buf , npthr ) ; vec = (float *)malloc(sizeof(float)*MAX(npthr,nathr)) ; for( iathr=0 ; iathr < nathr ; iathr++ ){ for( ipthr=0 ; ipthr < npthr ; ipthr++ ) vec[ipthr] = clust_thresh[ipthr][iathr] ; NI_add_column( nel , NI_FLOAT , vec ) ; } NI_set_attribute( nel , "commandline" , commandline ) ; NI_set_attribute( nel , "thresholding" , mlab ) ; sprintf(buf,"%d,%d,%d",nx,ny,nz) ; NI_set_attribute(nel,"nxyz",buf) ; sprintf(buf,"%.3f,%.3f,%.3f",dx,dy,dz) ; NI_set_attribute(nel,"dxyz",buf) ; sprintf(buf,"%.2f,%.2f,%.2f",fwhm_x,fwhm_y,fwhm_z) ; NI_set_attribute(nel,"fwhmxyz",buf) ; sprintf(buf,"%d",niter) ; NI_set_attribute(nel,"iter",buf) ; for( ipthr=0 ; ipthr < npthr ; ipthr++ ) vec[ipthr] = pthr[ipthr] ; nfar.num = npthr ; nfar.ar = vec ; bbb = NI_encode_float_list(&nfar,",") ; NI_set_attribute(nel,"pthr",bbb) ; NI_free(bbb) ; for( iathr=0 ; iathr < nathr ; iathr++ ) vec[iathr] = athr[iathr] ; nfar.num = nathr ; nfar.ar = vec ; bbb = NI_encode_float_list(&nfar,",") ; NI_set_attribute(nel,"athr",bbb) ; NI_free(bbb) ; if( mask_dset != NULL ){ NI_set_attribute(nel,"mask_dset_idcode",DSET_IDCODE_STR(mask_dset)) ; NI_set_attribute(nel,"mask_dset_name" ,DSET_HEADNAME(mask_dset)) ; sprintf(buf,"%d",mask_ngood) ; NI_set_attribute(nel,"mask_count",buf) ; } if( prefix != NULL ){ strcpy(fname,pname) ; strcat(fname,"niml") ; } else strcpy(fname,"stdout:") ; NI_write_element_tofile( fname , nel , NI_TEXT_MODE ) ; NI_free_element( nel ) ; if( prefix != NULL ){ if( refit_cmd == NULL ) refit_cmd = THD_zzprintf( refit_cmd , "3drefit " ) ; refit_cmd = THD_zzprintf( refit_cmd , "-atrstring AFNI_CLUSTSIM_NN%d_%s file:%s " , nnn , mlll , fname ) ; } if( prefix != NULL && mask_vol != NULL && first_mask ){ bbb = mask_to_b64string( mask_nvox , mask_vol ) ; first_mask = 0 ; if( bbb != NULL ){ FILE *fp ; sprintf(fname,"%s.mask",prefix) ; fp = fopen(fname,"w") ; if( fp != NULL ){ fprintf(fp,"%s\n",bbb); fclose(fp); } free(bbb) ; refit_cmd = THD_zzprintf( refit_cmd , "-atrstring AFNI_CLUSTSIM_MASK file:%s " , fname) ; } } } /* end of NIML output */ #if 0 switch( nnn*10 + mmm ){ case 11: clust_thresh_1sid_NN1 = clust_thresh ; break ; case 12: clust_thresh_2sid_NN1 = clust_thresh ; break ; case 13: clust_thresh_bsid_NN1 = clust_thresh ; break ; case 21: clust_thresh_1sid_NN2 = clust_thresh ; break ; case 22: clust_thresh_2sid_NN2 = clust_thresh ; break ; case 23: clust_thresh_bsid_NN2 = clust_thresh ; break ; case 31: clust_thresh_1sid_NN3 = clust_thresh ; break ; case 32: clust_thresh_2sid_NN3 = clust_thresh ; break ; case 33: clust_thresh_bsid_NN3 = clust_thresh ; break ; default: /* should never transpire */ ERROR_exit("nnn=%d mmm=%d :: This should never happen!",nnn,mmm) ; break ; } #else if( nnn*10+mmm == 11 ) clust_thresh_1sid_NN1 = clust_thresh ; if( nnn*10+mmm == 12 ) clust_thresh_2sid_NN1 = clust_thresh ; if( nnn*10+mmm == 13 ) clust_thresh_bsid_NN1 = clust_thresh ; if( nnn*10+mmm == 21 ) clust_thresh_1sid_NN2 = clust_thresh ; if( nnn*10+mmm == 22 ) clust_thresh_2sid_NN2 = clust_thresh ; if( nnn*10+mmm == 23 ) clust_thresh_bsid_NN2 = clust_thresh ; if( nnn*10+mmm == 31 ) clust_thresh_1sid_NN3 = clust_thresh ; if( nnn*10+mmm == 32 ) clust_thresh_2sid_NN3 = clust_thresh ; if( nnn*10+mmm == 33 ) clust_thresh_bsid_NN3 = clust_thresh ; #endif } /* end of loop over nnn = NN degree */ if( amesg != NULL ){ /* WARNING_message cannot handle "%s" with a very long string (>16K), so break this up a little 20 Feb 2014 [rickr] */ /* option: change to vsnprintf() in output_message()? */ WARNING_message("Simulation not effective for these cases:\n\n"); fprintf(stderr, "%s\n", amesg); fprintf(stderr, "*+ This means that not enough clusters, of any size, +*\n" " of voxels at or below each pthr threshold, were +*\n" " found to estimate at each alpha level. +*\n" "*+ In other words, the probability that noise-only +*\n" " data (of the given smoothness) will cause +*\n" " above-threshold (at the given pthr) clusters is +*\n" " smaller than the desired alpha levels. +*\n" "*+ This problem can arise when the masked region +*\n" " being simulated is small and at the same time +*\n" " the smoothness (FWHM) is large. +*\n" "*+ Read the 'CAUTION and CAVEAT' section at the end +*\n" " of the '-help' output for a longer explanation. +*\n\n"); free(amesg) ; amesg = NULL ; } } /* end of loop over mmm == sidedness */ if( do_athr_sum ){ int iathr ; FILE *fp ; char fname[256] ; thresh_summer_athr(athr_sum_bot,athr_sum_top) ; sprintf(fname,"%s.sumup.1D",(prefix!=NULL)?prefix:"ClustSim") ; fp = fopen(fname,"w") ; fprintf(fp,"# %s\n",commandline) ; fprintf(fp,"# " "1dplot -one -box -xaxis 0:0.1:10:5 -yaxis 0:0.32:16:2" " -xlabel 'nominal \\alpha at fixed p'" " -ylabel 'integrated FAR over p\\in[%g,%g]' " " -ynames 1s:NN1 1s:NN2 1s:NN3 2s:NN1 2s:NN2 2s:NN3 bs:NN1 bs:NN2 bs:NN3" " -x %s'[0]' -plabel '\\noesc %s' %s'[1..$]'\n" , pthr[athr_sum_top] , pthr[athr_sum_bot] , fname,fname,fname ) ; fprintf(fp,"# alpha ") ; fprintf(fp," 1s:NN1 1s:NN2 1s:NN3") ; if( do_athr_sum > 1 ){ fprintf(fp," 2s:NN1 2s:NN2 2s:NN3") ; fprintf(fp," bs:NN1 bs:NN2 bs:NN3") ; } fprintf(fp,"\n") ; for( iathr=0 ; iathr < nathr ; iathr++ ){ fprintf(fp," %.4f ",athr[iathr]) ; fprintf(fp," %.4f %.4f %.4f", rfa_1sid_NN1[iathr], rfa_1sid_NN2[iathr], rfa_1sid_NN3[iathr]) ; if( do_athr_sum > 1 ){ fprintf(fp," %.4f %.4f %.4f", rfa_2sid_NN1[iathr], rfa_2sid_NN2[iathr], rfa_2sid_NN3[iathr]) ; fprintf(fp," %.4f %.4f %.4f", rfa_bsid_NN1[iathr], rfa_bsid_NN2[iathr], rfa_bsid_NN3[iathr]) ; } fprintf(fp,"\n") ; } fclose(fp) ; } #ifdef USE_SHAVE destroy_shave() ; #endif #if 0 if( verb ) INFO_message("Clock time now = %.1f s",COX_clock_time()) ; #endif } /* end of outputizationing */ /*------- a minor aid for the pitiful helpless user [e.g., me] -------*/ if( refit_cmd != NULL ){ FILE *fp ; INFO_message("Command fragment to put cluster results into a dataset header;\n" " + (also echoed to file %s for your scripting pleasure)\n" " + Append the name of the datasets to be patched to this command:\n" " %s" , cmd_fname , refit_cmd ) ; fp = fopen(cmd_fname,"w") ; if( fp == NULL ){ ERROR_message("Can't write 3drefit command fragment to file %s :-(",cmd_fname) ; } else { fprintf(fp,"%s\n",refit_cmd) ; fclose(fp) ; } } /*-------- run away screaming into the night ----- AAUUGGGHHH!!! --------*/ exit(0); }