Fig. 3: Tumor detection accuracy of AI tool. | npj Precision Oncology

Fig. 3: Tumor detection accuracy of AI tool.

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

Fig. 3: Tumor detection accuracy of AI tool.The alternative text for this image may have been generated using AI.

Analysis includes biopsy slides with clear classification into benign or tumor-bearing (excluding 75 slides with “suspicious” regions, s. Fig. 1c). a Using maximal probability of being a tumor for different regions of single slides for identification of biopsy slides with tumor tissue. These thresholds were identified on a small internal validation dataset during algorithm development. b Using area threshold for identification of biopsy slides with tumor tissue. Confusion matrices for single slide AI-based classification compared to ground truth information: c Using a probability threshold, d Using a tumor area threshold. ACC overall accuracy, F1 F1 score, PPV positive predictive value, NPV negative predictive value, SENS sensitivity, SPEC specificity, AI: Ben slides classified as benign by AI tool, AI: Tu slides classified as tumor-bearing by AI tool, GT: Ben ground truth: benign slides, GT: Tu ground truth: slides containing tumor.

Back to article page