-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy paththoughts.txt
43 lines (38 loc) · 2.13 KB
/
thoughts.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
- Threshold data with value which maximizes coverage while minimizing noise
- Isolate a generous portion of the head scan which could contain the CoW
- Find the largest connected component which contains the voxels representing
the CoW
- accept multiple largest connected components if their total volumes are
relatively minimal
- Find medial axis
- Segement medial axis by identifying bifurcation points
- segement if the angle of bifurcation is greater than some threshold
for the average angle of the feeding artery
- if multiple elements satisfy previous condition, accept the path with
minimal angle
- Align medial axis mesh with a reference via ICP (PCA/SVD) alignment
- Eliminate paths in the medial axis graph which are not represented in the
reference medial axis graph
- Remove voxels which do not satisfy some minimal proximity threshold with the
updated medial axis graph
- Label remaining voxels by the corresponding medial axis path
- Correct artifacts caused by eliminating intermediate arteries (this is
probably very difficult)
Revised approach
Prerequisites: an averaged ideal CoW mesh with a corresponding medial axis map
with labeled segements
- Threshold and align iteratively until a desired level of correspondence or
path area is acquired OR prompt for manual threshold selection
- Isolate a generous portion of the thresholded mesh corresponding to the CoW
to speed up further computation
- Skeletonize mesh, and segment each path by angle deltas at each bifurcation
- Identify nearest neighbor segemented paths on the overlayed ideal medial axis
map, defining sufficient correspondence as a function of distance from the
reference and difference between total path lengths
- Remove all path segements from the target skeleton which do not meet some
correspondence threshold
- Traverse the updated skeleton and identify all voxels on the target mesh
which are contained within some radius of the skeleton (this radius could
be a function of the position on the skeleton for the reference mesh,
interpolated along the paths of the target skeleton)
- Remove all voxels which were not marked previously