Fig. 1: Slides from tissue microarrays (TMAs) with prostates samples from five sites were scanned, and the tissue regions were marked and extracted using QuPath (i.e., TMA slide image). | npj Imaging

Fig. 1: Slides from tissue microarrays (TMAs) with prostates samples from five sites were scanned, and the tissue regions were marked and extracted using QuPath (i.e., TMA slide image).

From: Artificial intelligence unravels interpretable malignancy grades of prostate cancer on histology images

Fig. 1

We then tiled each TMA core image into patches labeled by biochemical recurrence (BCR) status to develop our BCR model. We estimated the average BCR scores for each patient and applied survival modeling to introduce our novel risk-based grading for prostate cancer. The development set consisted of 600 patients, whereas the international external validation sets included three radical prostatectomy cohorts (CPCBN, PROCURE, and PLCO). The cohort description for all datasets included in this study can be obtained from Supplementary Tables S1–S3. PLCO: The Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. The endpoints we are shown in the black box. CSS: Cancer-specific survival. We emphasize that PC regions were manually demarcated on whole-slide images following the instruction given by a senior pathologist.

Back to article page