FIGURE 8 from A Pilot Study on Patient-specific Computational Forecasting of Prostate Cancer Growth during Active Surveillance Using an Imaging-informed Biomechanistic Model

Potential model-based biomarkers of higher-risk prostate cancer. A–F, Distribution of six potential model-based biomarkers in lower-risk (n = 8) and higher-risk (n = 8) prostate cancer, which were defined as tumors with GS = 3 + 3 and GS ≥ 3 + 4, respectively. These model-based biomarkers were calculated at the times were both a histopathologic assessment and imaging measurement are available for each patient in the cohort (n = 7). In particular, the model-based markers in A–F are: prostate volume (VP), tumor volume, total tumor cell volume (VN), mean normalized tumor cell density (), total tumor index (NT), and mean proliferation activity of the tumor (Ap). Outliers are represented as hollow circles and an asterisk indicates significance under a one-sided Wilcoxon rank-sum test (P G, ROC curves for (i) the univariate logistic regression model constructed using the mean proliferation activity of the tumor (i.e., the only model-based marker that was significantly different between lower-risk prostate cancer and higher-risk prostate cancer), and (ii) the bivariate logistic regression model constructed using the mean proliferation activity of the tumor and the total tumor index (which is the combination of model-based markers that rendered the highest performance). The AUC of each ROC curve is reported within the plot, and the optimal performance point for both the univariate and bivariate logistic regression models operates at 75% sensitivity and specificity (bullet points on the ROC curves).