Hello,

I'm wondering regarding the inputs and equation for the network construction.

In the NeuroImage paper,
the weights of network was defined as "n(j) / n(tot)"
when "n(j) = number of fibers passing through region j"
  and "n(tot) = total number of fibers from seed mask i".

How can I get "n(j)" and "n(tot)"? 
I tested in MATLAB as below.

fdt_paths.nii.gz: result of probtrackx with seedmask i
Label03_SeedMask.nii.gz: seedmask j (in here, j = 3)

fdt = read_avw("fdt_paths.nii.gz");
roi_j = read_avw("Label03_SeedMask.nii.gz");
roi_loc = find(roi_j>0);

(1) divided by sum of fdt_paths
w = sum(fdt(roi_loc)) / sum(fdt(:))
(result = 0.0043)

(2) divided by waytotal
w = sum(fdt(roi_loc)) / waytotal
(result = 0.0374)

However, "SBJID_ProbabilisticMatrix_OPD_90.txt" reported 0.0187 at row=1, col=3; and 0.0445 at row=3, col=1.
I understood that row indicates the seed index and column indicates target index.

How could I calculate network matrix manually?

Thank you for your advances,

YH

Hi Kim,

Thanks for your interest in PANDA.
It is normal that the probability at row=1, col=3 is different from the probability at row=3, col=1.

The network matrix of the probabilistic tracking is asymmetric.
During the probabilistic tracking, each step selects a direction with a probability. The path of the fibers will be always different when you select the ROI1 and ROI3, because starting from the seed, every the direction of each step has a degree of randomness. 
For example, if you select point A in ROI1 as seed and the fiber end in point B in ROI3, the fiber may not end in point A in ROI1 when you select point B in ROI3 as seed.

Generally, we always sum the probability at row=1, col=3 and the probability at row=3, col=1 and then divide 2, and then treat the result as the probability of the presence of the fiber between ROI1 and ROI3.

Best

Zaixu