TractX: filtering and validation questions

pretty tractogram ≠ correct connections, annoying but useful reminder
keep validity definition with the score. especially when changing clustering / endpoint voting

problem: a streamline's visual plausibility is not enough to establish that it corresponds to a valid connection. Keep tractogram generation, filtering, bundle recognition and validation as separate stages.

Five-step tractogram validation procedure.

notebook evaluation outline: identify valid connections using bundle clustering and robust shape; classify remaining tracks as invalid connections; cluster invalid connections with RecoBundlesX; assign best-matching endpoint regions by majority vote; merge bundles with similar termination regions. The notebook later records overlap and overreach alongside precision and valid-connection ratio.

reference: the journal cites the tractography validation study. Revisit the metric definitions before comparing scores from different filtering criteria.

• review questions: which reference labels define validity? Does a reduction in overreach discard valid coverage? Keep subject split, ground truth and metric implementation attached to every result. The journal's broad “gain over state of the art” statement is not enough to reproduce a benchmark comparison.