Researchers at Google have built a model that takes a high resolution image of prostate tissue and predicts the severity of cancer present. It does this by splitting the image into thousands of tiny square patches and assigning each a Gleason pattern: benign, pattern 3 (cancerous, but might not require intervention), and the more advanced pattern 4 and pattern 5.

To explore the use of model assistance in pathology, pathologists use a fullscreen version of this interface while grading tissue slides. Patches with the same pattern are grouped together and given a colored outline.