Skip to main content
. Author manuscript; available in PMC: 2020 Oct 1.
Published in final edited form as: J Neurosci Methods. 2019 Aug 1;326:108373. doi: 10.1016/j.jneumeth.2019.108373

Table 2 -.

Total ROI counts at different stages of segmentation workflow. Thresholding and assigning of ROIs at axon myelin boundary is done on preliminary segmented images, Interim dataset, for control purposes. Improved segmentation reduces the number of ROIs in white matter images. After measurements by the ray measurement tools (RMT), a final cleanup of non-fiber ROIs and notably damaged fibers is obtained by suitably formulated conditions on exit from the RMT. See Supplementary Information 3, for specification of additional conditions and examples of spurious ROIs. Counts for all three image sets were obtained using deep neural network tool trained only on 20 images from WM_Set_01.

ROI number / Image set WM_Set_01 WM_Set_02 ON_Set_01 comment
ROI num. at INTERIM stage 11,277 10,977 3,932 After pre-segmentation and post-processing
ROI num. in DNN segmented 7,897 9,379 6,762
Num. ROI discarded in RMT by ray measurements 522 1,661 2,122 ROIs are discarded due to no usable rays
Num. ROI discarded in RMT by add. conditions 1,421 1,370 2,017
Final ROI num. after RMT 5,954 (7,635)* 6,347 (7,851)* 2,623 (3,000)* * remaining ROIs when RMT is applied at INTERIM
ROI num. CORRECTED data 5,945 6,168 2,775 ROIs in the annotated set