Computer Vision News - September 2020

Medical Imaging 10 Paige is a start-up using cutting-edge artificial intelligence to transform the diagnosis and treatment of cancer. It spun out of the Memorial Sloan Kettering Cancer Center (MSK) in 2017 with the aim to bring computational pathology solutions to the market. We speak to its Chief Business Development Officer, Carla Leibowitz, about its first three years and what the future holds. Computational pathology is the application of machine learning techniques to pathology. Where before pathologists would look at images under a microscope, now digital pathology scanners digitize those images and combined with additional data they can be used to build incredible machine learning systems. Carla tells us that Paige has three unique pillars that set it apart from other companies in the space. The first is proprietary machine learning techniques . Its system takes tens of thousands of very large pathology images – can reach above billion pixels and four gigabytes each – and trains on them at scale, without needing pixelwise annotation or hordes of people curating findings. In many cases, the pathologist will have reviewed several images for a single case and issued a report – that report is tied to each image or set of images and the system learns from that. Its second unique advantage is an exclusive license to use the pathology slides at MSK for commercial purposes . It has permission to digitize the image archive there, except for cases where patients did not consent, which contains cases that have been diagnosed by some of the world’s best pathologists, so the ground truth is highly accurate. MSK can use the images for research purposes, so it benefits as well. Paige can also request additional information such as outcomes data and genomics profiles for those patients. And the third pillar? “Definitely the team ,” Carla declares with pride. “Our founder, Professor Thomas Fuchs , is the father of computational pathology, having Carla Leibowitz