Deep learning models trained on patches from whole slide images can match pathologist performance in detecting cancer. A 2017 study used 270 whole slide images to train AI that identified cancerous tissue as accurately as human experts.
270 whole slide images.
Deep learning has had an incredible impact on medical imaging. For diagnosis, one of the tasks that pathologists do is look at scanned microscopic images of tissue called HOL slide images, and determine the extent to which a cancer has spread. This is important to help plan treatment, predict the course of the disease, and the chance of recovery. In one study in 2017, using only 270 whole slide images images, AI algorithms were developed and then evaluated against pathologists and it was found that the best algorithms performed as well as the pathologists did. The general setup of these studies is that instead of feeding in one large, high resolution digital image of the slide, several patches are extracted at a high magnification and used to train a model. These patches are labeled with the original label of the whole slide image and then fed into a deep learning algorithm.
