The paradox of medical AI implementation

In 2012, the era of deep learning AI began with the convolutional neural network (AlexNet), which won the ImageNet Challenge. These images were of everyday objects, animals and scenes that had nothing to do with health and medicine. I wrote a review over 7 years ago Natural medicine justified High-performance medicine It summarized the remarkable advances in AI interpretation of medical images. Almost every type of medical image has now undergone comprehensive assessment with AI, including X-ray, CT, MRI, ultrasound, pathology slides, skin anomalies, electrocardiograms, endoscopy, and retinal photos. A few weeks ago here Basic truths and then in The lancet I’ve written about three AI tools that should be used with every mammogram, based on the largest randomized trial of more than 100,000 women and two recent FDA approvals. There were 44 randomized trials on colonoscopy which consistently and overall demonstrate a significant advantage of AI support in detection…