Table 5 The patch number and the cancer proportions in the external datasets curated from 60 whole-mount histology images.

From: Critical evaluation of artificial intelligence as a digital twin of pathologists for prostate cancer pathology

External dataset for model comparison (60 whole mount images)

Finding

Patches

Noncancer tissues, n (%)

142,135 (81.95)

Prostate Cancer, n (%)

31,286 (18.05)

Total, n (%)

173,421 (100.00)

Average pixelwise cancer proportion in a patch labeled with prostate cancer (median)

80.6% (100%)

  1. One whole-mount image covers the whole prostatic slice and provides patches on average 6 times more than a single whole-slide image of the Cancer Genome Atlas for prostate cancer.
Back to article page