TractX metrics: definitions, targets and conflicting poster arithmetic

same 23→72.83 endpoints, two different percentage-gain claims in the posters. ugh
recompute from the raw numbers; keep scoring definition + denominator nearby. percentage points and percent gain are not the same thing

• Engineering-plan targets include at least 21 of 25 valid bundles, streamline VC above 54%, fewer than 88 invalid bundles, streamline IC below 36%, and bundle precision above 23%. These are proposed criteria, not results. The plan names a 2021 M1 iMac with 16 GB RAM, possible cloud computation and a $100 budget. Its body date says January 9, 2023 while references extend into January 2024, so leave the exact historical date unresolved.

• Later poster values are VC/recall 94.52%, precision 72.83%, overlap 79% and overreach 19%. Its evaluation describes an alternative “valid connection” definition, bundle clustering/shape checks, RecoBundlesX clustering of remaining ICs, majority-vote endpoint regions and merging similar terminations. Keep that evaluation definition attached to the numbers rather than assume identical scoring to a published baseline.

• There is a real source discrepancy: Poster Materials says a 216.15% precision gain, while Poster Board (Combined) says 116.15%, despite both quoting 23% → 72.83%. Direct arithmetic gives a 49.83 percentage-point increase or about 216.65% relative to exactly 23%. A more precise unrounded baseline might alter the latter slightly, but it cannot justify silently merging the conflicting versions. Retain the raw endpoint values and ask for the original evaluation table.

• The poster's normalized confusion matrix has its own 72.84% value and (n=33) label; the speaking notes mention five repetitions. Without run identifiers and clear denominators, do not combine these into one sample size. No independent reproduction or clinical validation is established by this document set.

TractX output in three orthogonal views.

TractX output in a slice lightbox.