I am reviving this thread because 3dROIstats is behaving weirdly using my surface parcellation. I think I've found a solution, but also maybe a bug?

(In narrator voice) Previously on this thread:

I converted a Freesurfer annotation to .niml.dset using FS_read_annot, then I resample it to -ld 60 and ranked the node values so that 3dROIstats could handle the parcel labels.

---

I am now extracting mean time series from each parcel using time series projected to the surface and the results from 3dRank as the mask.

I have examined the .rankmap.1D file and the mapping from rank to annotation values seems to be in order and I have the correct number of labels. However, when extract the time series using the ranked parcellation as a mask, I'm getting weird results.

What I expect: I have 201 labels, 200 parcels of interest + the medial wall for each hemisphere. I was expecting 201 columns of timeseries from 3dROIstats with labels of Mean_0 to Mean_200, which should correspond to the rank values in .rankmap.1D

What I'm getting: 200 columns with the following values:

Mean_1  	Mean_2  	Mean_3  	Mean_4  	Mean_5  	Mean_6  	Mean_7  	Mean_8  	Mean_9  	Mean_10  	Mean_11  	Mean_12  	Mean_13  	Mean_14  	Mean_15  	Mean_16  	Mean_17  	Mean_18  	Mean_19  	Mean_20  	Mean_21  	Mean_22  	Mean_23  	Mean_24  	Mean_25  	Mean_26  	Mean_27  	Mean_28  	Mean_29  	Mean_30  	Mean_31  	Mean_32  	Mean_33  	Mean_34  	Mean_35  	Mean_36  	Mean_37  	Mean_38  	Mean_39  	Mean_40  	Mean_41  	Mean_42  	Mean_43  	Mean_44  	Mean_45  	Mean_46  	Mean_47  	Mean_48  	Mean_49  	Mean_50  	Mean_51  	Mean_52  	Mean_53  	Mean_54  	Mean_55  	Mean_56  	Mean_57  	Mean_58  	Mean_59  	Mean_60  	Mean_61  	Mean_62  	Mean_63  	Mean_64  	Mean_65  	Mean_66  	Mean_67  	Mean_68  	Mean_69  	Mean_70  	Mean_71  	Mean_72  	Mean_73  	Mean_74  	Mean_75  	Mean_76  	Mean_77  	Mean_78  	Mean_79  	Mean_80  	Mean_81  	Mean_82  	Mean_83  	Mean_84  	Mean_85  	Mean_86  	Mean_87  	Mean_88  	Mean_89  	Mean_90  	Mean_91  	Mean_92  	Mean_93  	Mean_94  	Mean_95  	Mean_96  	Mean_97  	Mean_98  	Mean_99  	Mean_100  	Mean_101  	Mean_102  	Mean_103  	Mean_104  	Mean_105  	Mean_106  	Mean_107  	Mean_108  	Mean_109  	Mean_110  	Mean_111  	Mean_112  	Mean_113  	Mean_114  	Mean_115  	Mean_116  	Mean_117  	Mean_118  	Mean_119  	Mean_120  	Mean_121  	Mean_122  	Mean_123  	Mean_124  	Mean_125  	Mean_126  	Mean_127  	Mean_molecular_layer_HP  	Mean_129  	Mean_130  	Mean_131  	Mean_132  	Mean_133  	Mean_134  	Mean_135  	Mean_136  	Mean_137  	Mean_138  	Mean_139  	Mean_140  	Mean_141  	Mean_142  	Mean_143  	Mean_144  	Mean_145  	Mean_146  	Mean_147  	Mean_148  	Mean_149  	Mean_150  	Mean_151  	Mean_152  	Mean_153  	Mean_154  	Mean_155  	Mean_156  	Mean_157  	Mean_158  	Mean_159  	Mean_160  	Mean_161  	Mean_162  	Mean_163  	Mean_164  	Mean_165  	Mean_166  	Mean_167  	Mean_168  	Mean_169  	Mean_170  	Mean_171  	Mean_172  	Mean_173  	Mean_174  	Mean_175  	Mean_176  	Mean_177  	Mean_178  	Mean_179  	Mean_180  	Mean_181  	Mean_182  	Mean_183  	Mean_184  	Mean_185  	Mean_186  	Mean_187  	Mean_188  	Mean_189  	Mean_190  	Mean_191  	Mean_192  	Mean_193  	Mean_194  	Mean_195  	Mean_196  	Mean_197  	Mean_198  	Mean_199  	Mean_200

Note that there is no Mean_0, which I guess makes sense. Oddly though, look at what should be Mean_128. It says Mean_molecular_layer_HP. I have no idea where that came from because the node values are stored as float:

Number of values stored at each pixel = 1
  -- At sub-brick #0 'rank' datum type is float

As a solution, I am using 3dcalc to just add 1 to all nodes. This gives me the expected result of Mean_1 to Mean_201 and I can just subtract 1 to map it back to the annotation value. I just wanted to bring this up because I can't imagine that this is the intended behavior of 3dROIstats. I understand that it skips over nodes with a 0 value, but why Mean_molecular_layer_HP?