Additional file 1 of Whole anterior visual pathway segmentation from high-resolution MRI using artificial intelligence

Résumé fourni par la source

Additional file 1: Table S1. Summary of all spatial similarity metrics between Reader 1 (R1) and Reader 2 (R2) for the merged aVP ground-truth masks. Table S2. Summary of all spatial similarity metrics between A.I. and GT, each aVP ON, OC and OT segment. Higher values for Dice, Jaccard, Precision, Recall, and Volumetric similarity, and lower values for Hausdorff, HD95, and Average surface distance, indicate higher morphology similarity. Figure S1. Common anatomical pitfalls in the segmentation of the aVP. Left: Intracanalicular segment CISS images (coronal, axial, sagittal) showing reduced contrast between the optic nerve and surrounding cerebrospinal fluid, leading to boundary uncertainty. Right: Optic tracts example from subject 0035 illustrating high inter-reader variability in defining the posterior truncation of the tract. Yellow ellipses indicate regions of disagreement between readers. Figure S2. Representative axial, coronal, and sagittal CISS images from an external subject acquired at a different center using a different scanner with the same acquisition protocol are shown, with the automated aVP-seg output overlaid. The segmentation delineates the optic nerves, chiasm, and optic tracts using color coding, together with the corresponding 3D volume rendering. This example is provided for illustrative purposes only; no quantitative evaluation was performed, and no conclusions regarding cross-scanner or cross-protocol generalizability can be drawn from a single case.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.