AI predicts breast cancer and eliminates racism

MIT Computer Science and Artificial Intelligence Research Institute CSAIL and Massachusetts General Hospital research team developed a new breast cancer x-ray test using AI. AI trained with a deep learning model retrieves mammograms that are likely to develop breast cancer within 5 years. The models developed so far have different detection levels for whites and blacks, but there is no such difference in new models.

The research team first selected more than 60,000 mammograms of patients treated at the Massachusetts General Hospital and selected pictures of breast cancer within 5 years. The data was then used to train AI to recognize the subtle degeneration of breast tissue that shows early signs of breast cancer. As a result, only 18% of patients who should be classified as the highest risk were predicted by the existing method, but the new model showed excellent prediction accuracy at 31%.

This AI could also help bridge the racial gap in healthcare. There have been reports that the detection of women of color is delayed because the early detection model for breast cancer is mainly based on data on white women. Black women have a 43% higher rate of breast cancer mortality than white women. Black, Hispanic and Asian women are also younger than white women.

The research team confirmed that AI was able to detect breast cancer uniformly regardless of race with a simple factor of learning mammograms of all races. The research team said that the new AI model achieved high detection precision regardless of race, as a remarkable achievement in this study, and said that if the model is used extensively in the future, it could significantly improve the current way of estimating patients. The research team hopes to apply this model to low-risk models of other diseases, diseases, and pancreatic cancer. Related information can be found here .