Everyone knows that MRI is better than mammography at diagnosing breast cancer. Because of its expense and resource constraints, it is not practical for all women to receive MRI screening. So who should get an MRI? Researchers in Sweden have developed artificial intelligence software to read mammograms in tandem with radiologists. Using this technology they identified 6.9% of women who had just gone mammography with adverse AI scores based on the use the AI to reread the mammograms for further screening with MRI, and randomized them to receive or not receive MRI, then compared results. Their report is available online free of charge from the journal Nature Medicine. Click here to be taken to the article. They looked at the outcomes of both groups and concluded that those women with adverse AI scores who underwent mammography were four times as likely to have breast cancer diagnosed as those women (not in the study) who underwent MRI without pre-selection(64 cancers vs. 16 per 1000 recipients of MRI). All of the women in the study would have had their diagnosis delayed without MRI.
So what does this mean? Physicians worry what role AI will have in reshaping the practice of medicine. This study shows that AI can improve outcomes without interfering with the practice of medicine.