Pathology AI Models Show Demographic Bias in Cancer Diagnosis
Researchers at Harvard Medical School found that four standard AI pathology models used for cancer diagnosis exhibit demographic bias, performing unequally across populations based on gender, race, and age. The team developed FAIR-Path, a framework that successfully reduced bias in the models, and identified the need for systematic bias testing in pathology AI to ensure equitable patient care.
What happened
Researchers at Harvard Medical School investigated bias in four standard AI pathology models designed for cancer diagnosis by analyzing them against a large, multi-institutional repository of pathology slides spanning 20 cancer types.
All four models demonstrated unequal performance in detecting and differentiating cancers across populations based on patients' self-reported gender, race, and age.
Reported impact
The researchers found that pathology AI models can infer demographic information from pathology slides and exhibit bias in cancer diagnosis among different populations.
According to the study, identifying and counteracting AI bias in medicine is critical because it can affect diagnostic accuracy and patient outcomes.
Index analysis
The researchers identified three possible explanations for the demographic bias and developed a framework called FAIR-Path that helped reduce bias in the models.
FAIR-Path's success indicates that researchers can improve the fairness of AI models for cancer pathology with minimal effort.
Governance controls
The findings highlight the need to systematically check for bias in pathology AI to ensure equitable care for patients.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.