When using BAFET to analyze brain vessels, I manually add tags to each image set based on subjective feeling, such as "Many" for rich vessel in brain and "Few" for few vessels connecting to main arteries. I want to make validation between human feeling and statistical data from feature detected so that the result from features is discriminative and compelling.

All the tags I marked:

Few: brain with evidently fewer vessel branches, especially Order 6+ branches. Length of end branches is mostly smaller than normal.

Many: brain with evidently more vessel branches, especially Order 6+ branches. Length of end branches is mostly bigger than normal.

Complete: A complete structure of Circle of Willis is found, even the vessel in the circle is thin and the total number of branches is smaller than normal.

Typical: The structure of brain vessels is ideally normal, with median number of branches and normal structure. Such case is clear enough to be used as demonstration.

Strange: Some vessels are strange in location or appearance, which is unsuitable to compare with other cases.

1. Validation on vessel appearance in richness

Compare tags between "Few" and "Many".

All cases marked as "Many" has longer total vessel and "Order 6+" vessel (small vessels in the end branches). In the comparison between average total vessel length, cases marked as "Many" (mean value 4442.028) is twice of those marked as "Few" (mean value 2098.074); In the comparison between average Order 6+ vessel length, cases marked as "Many" (mean value 3082.164) is six times of those marked as "Few" (mean value 459.536)

The result means the number of length extracted from BAFET is in accordance with visual feeling, so the length of vessel given by BAFET is reliable in analyzing the richness of brain vessel.

Total vessel length of brain between the group taged as "Few" and "Many"

Order 6+ vessel (small vessels in the end branches) length of brain between the group taged as "Few" and "Many"

2. "Strange" tags

Most cases are similar in direction of arteries. Similar structures make it comparable to each other in their length, volume and other features. However, some of the cases are very special in vessel structure, which makes them difficult to compare features with others.

In one of the cases, one main artery is linked to front brain and the other one linked to behind. So the total length, volume and other features are largely different from other cases. In our research, only two main arteries are used, so manual modification to strange structures are needed to make each case similar and comparable.

3. "Typical" tags typical in number?

Typical cases are similar both in number and in appearance. The span of total length is 1326.29, much smaller than group span 5586.78. The span of 4+ length is 1179.85, much smaller than group span 4615.47.

Total vessel length of brain between four groups

Order 6+ vessel (small vessels in the end branches) length of brain between four groups