IBM bringing its skin-cancer computer vision system to hospitals

IBM says it has developed a machine learning system that identified images of skin cancer with better than 95 percent accuracy in experiments, and it’s now teaming up with doctors to see how it can help them do the same. On Wednesday, the company announced a partnership with Memorial Sloan Kettering — one of IBM’s early partners on its Watson system — to research the computer vision technology might be applied in medical settings.

According to one study, cited in the IBM infographic below, diagnostic accuracy for skin cancer today is estimated at between 75 percent and 84 percent even with computer assistance. If IBM’s research results hold up in the real world, they would constitute a significant improvement.

As noted above, the skin cancer research is not IBM’s first foray into applying machine learning and artificial intelligence techniques — which it prefers to call cognitive computing — in the health care setting. In fact, the company announced earlier this week a partnership with the Department of Veterans’ Affairs to investigate the utility of the IBM Watson system for analyzing medical records.

And [company]IBM[/company] is certainly not the first institution to think about how advances in computer vision could be used to diagnose disease. Two startups — Enlitic and Butterfly Network — recently launched with the goal of improving diagnostics using deep learning algorithms, and the application of machine learning to medical imagery has been, and continues to be, the subject of numerous academic studies.

We will be discussing the state of the art in machine learning, and computer vision specifically, at our Structure Data conference in March with speakers from IBM, Facebook, Yahoo, Stanford and Qualcomm, among others.