AI Developed at Memorial Sloan Kettering Cancer Center with Nearly 100% Detection Rate

Jul 16, 2019 | AI, Cancer, Cancer Detection, Pathology

White robot hand scanning human body 3D rendering

Memorial Sloan Kettering Cancer Center has developed a clinical-grade artificial intelligence model for pathology that can detect cancer in tissue slides with near 100% accuracy.

The Study

The Memorial Sloan Kettering Cancer Center-led team recently published the study results in Nature Medicine.

During the study, New York preeminent institution’s team analyzed slides from more than 15,000 prostate, skin, and breast cancer patients. The Artificial Intelligence (AI) had the ability to detect cancer with 100% sensitivity and even was able to preclude most cancer negative samples. This sped up the ability of pathologists to focus in on only the most critical slides for malignant cancer.

Weakly Supervised AI

Unlike many other AI models, the Memorial Sloan Kettering Cancer Center AI is “weakly supervised,” meaning the research team trained the machine on a massive set of only nominally annotated slides instead of a small sample of slides with extensive annotation. The team believes that this model better account for variability between the various forms of cancer.

Lead Research/Investigator

Thomas J. Fuchs, Department of Pathology, Memorial Sloan Kettering Cancer Center


Pin It on Pinterest