Skip to main content
. Author manuscript; available in PMC: 2022 Feb 2.
Published in final edited form as: IEEE Trans Med Imaging. 2021 Feb 2;40(2):635–647. doi: 10.1109/TMI.2020.3034038

Fig. 6.

Fig. 6.

Top panel shows invalid (IC) and valid (VC) connection rate of each submission made to the ISMRM 2015 challenge. Colors indicate different teams [9]. Using the same color coding, left and right panels on the second row show comparisons for numbers of invalid (IB) vs. valid (VB) bundles, and bundle overreach vs. overlap rates obtained with different submissions. The eleven tested parameter combinations, varied around a default setting are shown on the right. Increasing fmin values decreases bundle overlap, overreach and IB as expected. Our approach was able to capture all fiber bundles except posterior commissure which was not observed in any of the original submissions either. Our approach scored well-over 60% overlap while keeping the overreach just below 40% which was not achieved by any of the original submissions. In general, compared to others, we observe that our results score the minimum overreach for any given bundle overlap. We observe that decreasing the step size decreases overlap, overreach and IC. Increasing probe radius and length improves VC, IC and bundle overreach while they compromise from bundle overlap.