Fig. 6.

Individual visualization of the combined model through SHAP. (A) and (B) show two examples of correctly predicted csPCa cases. (C) and (D) show two examples of correctly predicted non-csPCa cases. (A) for this specific sample, the PI-RADS score, Cit level, peri-rad-score, and intra-rad-score increase the prediction probability of csPCa, whereas the PSAD and PSA have the opposite effect. (B) exhibits a similar pattern. In contrast, (C) shows that the peri-rad-score, PSAD, and PSA increase the predicted probability of non-csPCa, whereas the intra-rad-score and PI-RADS score have the opposite effect. Figure (D) displays a comparable pattern to (C)