Samseg Aseg Test | March 6 2017

Andrew ran samseg on the ADNI60 test set. mri_compute_seg_overlap was used to compare the subject's segSubSpace.mgz with its aseg.mgz.

results

The mean overlap of the majority of subjects was above 0.8, but 19 subjects had a mean below 0.2

subjs_dice.png
fig 1: mean dice scores for all ADNI60 subjects

The low dice scores seen in each failed subject were caused by inaccurate registration:

0341_regerror.png0123_regerror.png0934_regerror.png
fig 2: subjects 0341, 0123, and 0934 with failed segSubSpace registration

With the failed outliers removed, the mean dice coefficients for each label appear better than those in the previous buckner40 test (SamsegEvaluationMarch2016)

labels_dice.no_outliers.png
fig 3: mean dice scores for each label over all non-outlier ADNI60 subjects

data location

main test directory: /autofs/cluster/fsm/users/samseg/tests/2017-03-06_ADNI60

charts and snapshots: /autofs/cluster/fsm/users/samseg/tests/2017-03-06_ADNI60/analysis

adni test set data: /autofs/cluster/fsm/users/samseg/subjects/ADNI60

SamsegAsegTesting/2017-03-06_ADNI60 (last edited 2017-03-17 10:45:26 by AndrewHoopes)