Fig. 1: Principle of AI tool, study design, and characteristics of study cohorts. | npj Precision Oncology

Fig. 1: Principle of AI tool, study design, and characteristics of study cohorts.

From: An international multi-institutional validation study of the algorithm for prostate cancer detection and Gleason grading

Fig. 1

a The AI tool consists of tissue detection-, prostate cancer detection-, and Gleason grading-modules representing different deep learning-based algorithms. The prostate cancer detection module also detects other tissue classes, such as benign glandular, stromal tissue, high-grade prostatic intraepithelial neoplasia (HGPIN) and some others. b Study design includes validation of the AI tool using material from five pathology departments. Two cohorts (tumor-bearing slides) were used for validation of Gleason grading and were analyzed by 11 board-certified pathologists and AI tool. c Slides from five departments were included in the study. WNS B represents a subcohort of WNS A scanned by a different histoscanner. #A negligibly small subset of temporally separated biopsy slides (UKK-1) was originally included into the training dataset. We provide this information for transparency. *This calculation excludes UKK-1 slides. UKK University Hospital Cologne, WNS Hospital Wiener Neustadt, TRO Institute of Pathology Troidorf, ACH University Hospital Achen, BRA Municipal Hospital Brunswick.

Back to article page