Pathology AI models show demographic bias in cancer diagnosis

Pathology AI models show demographic bias in cancer diagnosis

Pathology has long been the cornerstone of cancer diagnosis and treatment. A pathologist carefully examines an ultrathin slice of human tissue under a microscope for clues that indicate the presence, type, and stage of cancer.

What AI Learned From Cancer Slides Shocked Researchers

AI can spot cancer—but it can also spot who you are, and that turns out to matter. A new study finds that artificial intelligence systems used to diagnose cancer from pathology slides do not perform equally for all patients, with accuracy varying across race, gender, and age groups. Researchers uncovered three main reasons behind this [...]

One in Three AI Cancer Diagnoses Vulnerable to Demographic Bias

Artificial intelligence systems for diagnosing cancer from pathology slides often vary in accuracy according to a patient’s age, race and sex, researchers report. The findings, in Cell Reports Medicine, stress the need to check for bias in these data-driven algorithms to ensure they enhance diagnostic accuracy even when overall performance appears strong. The AI issues may have been due to underrepresented patients in training sets, differences …