Skip to main content
. Author manuscript; available in PMC: 2024 Aug 1.
Published in final edited form as: J Pathol. 2023 May 26;260(4):390–401. doi: 10.1002/path.6090

Figure 4. Clinical analysis comparing the prognostic value of 3D & 2D nuclear features.

Figure 4.

(A) ROC curves are shown for multiparameter models trained on 3D (red) and 2D (blue) nuclear features in epithelial regions. (B) ROC curves of multiparameter models trained on 3D (red) and 2D (blue) nuclear features in stromal regions. (C,D) Kaplan–Meier curves are shown for BCR-free survival, showing that the model trained on epithelial 3D nuclear features (C) can better stratify patients into low- and high-risk categories than the model trained on epithelial 2D nuclear features (D). (E,F) Violin and box plots are shown for two examples of epithelial 3D nuclear features, along with their analogous 2D features, for cases in which BCR occurred within 5 years of radical prostatectomy (“BCR”) and for cases in which there was no BCR within 5 years of radical prostatectomy (“non-BCR”). For both example features, “Mean epithelial nucleus-to-convex-hull ratio” in (E), and “epithelial nucleus-to-convex-hull variance” in (F), the 3D feature shows improved patient risk stratification compared to its 2D counterpart.