Extended Data Fig. 7: Evaluation of the robustness of Gleason Morisita index.
From: Tumor evolution metrics predict recurrence beyond 10 years in locally advanced prostate cancer
To evaluate the robustness of Gleason Morisita index, two comparisons are made: balance of epithelial cells vs. balance of segmented pixels (A, B), and Voronoi sub-regions vs. rectangular subregions (C, D). (A) Scatter plot comparing patient-level Gleason Morisita indices from the cell and segmentation-based metrics. The two metrics are well correlated, with a Pearson correlation of 0.82 (Imaging Cohort, n = 250 participants, d.f. = 248, P = 5.17 × 10−62). (B) KM curve of time to recurrence for segmentation-based Gleason Morisita index, split by median. Segmentation-based metric is also a significant predictor of time to recurrence (Imaging Cohort, n = 250 participants, two-sided log-rank test, χ2 = 10.43, d.f. = 1, P = 0.00039), with the pattern of survival closely resembling that of the cell-based metric (Fig. 4D). (C) Scatter plot comparing patient-level Gleason Morisita indices from the Voronoi and rectangular regions. The two metrics are well correlated, with a Pearson correlation of 0.86 (Imaging Cohort, n = 250 participants, d.f. = 248, P = 4.3 × 10−73). (D) KM curve of time to recurrence for Gleason Morisita index from rectangular regions, split by median. Rectangular region metric is also a significant predictor of time to recurrence (Imaging Cohort, n = 250 participants, two-sided log-rank test, χ2 = 5.94, d.f. = 1, P = 0.0035), with the pattern of survival closely resembling that of the Voronoi region metric (Fig. 4D).
