How Deep Learning Could Catch Breast Cancers that Humans Miss, Blockchain Used to Store and Share Data
On the Nvidia website, a post by James Beckett notes that ” mammograms save lives by detecting breast cancer early, except when they don’t.” https://blogs.nvidia.com/blog/2018/01/25/ai-improve-mammogram-accuracy/ Dr. Dexter Hadley is a professor of pediatrics, pathology, and laboratory medicine with the Institute for Computational Health Sciences at the University of California, San Francisco. Hadley, who is an engineer as well as a physician, is working to change that. He and his colleagues at UCSF are using GPU-accelerated deep learning to improve mammographic accuracy. https://www.nvidia.com/en-us/deep-learning-ai/
To improve mammogram accuracy, the UCSF researchers are trying to train a neural network to identify cancers at an earlier stage. It was also noted that more than half of women who undergo screening over a 10 year period experience a false positive resulting in call-backs and biopsies, all with ensuing stress.
To solve the problem, the UCSF team took over 30,000 written pathologist reports and used deep learning algorithms to infer how individual cancer patients fared. The researchers linked this data to more than 700,000 mammogram images. Then, using a proprietary deep learning framework, they trained a neural network to predict cancer diagnoses based on breast imaging.
Hadley’s next move is to get access to the nearly 40 million mammograms performed yearly in the U.S. Currently privacy laws (HIPPA) restrict his access. Consequently, Hadley and his partners have put out calls to millions of women to donate their mammograms or other imaging data for research. Hadley’s hoping to obtain 5 million mammograms so that artificial intelligence can develop a robust model. Women can sign up to participate at https://www.breastwecan.org/
Hadley and his colleagues are building a system that allows people to share their medical data with researchers easily and securely — and retain control over it. Their method, which is based on blockchain technology that underlies all of the cryptocurrencies including Bitcoin. https://www.nature.com/articles/d41586-018-02641-7?ito=792 By May, Hadley and his colleagues will launch a study to train their artificial intelligence algorithm to detect cancer using mammograms that they hope to obtain from between three million and five million US women. The UCSF team joins a growing number of scientists who are using who are using blockchain technology to share medical records in a more secure and efficient manner. Companies are now appearing in the marketplace to broker data exchanges between individuals, companies, and researchers. A blockchain system is currently being tested at Beth Israel Deaconess Medical Center in Boston this year. Ultimately for this to work people will need to share their data with researchers and clinicians. In a blockchain ledger, only the patient has access to their data unless permission is given to share. Might this be the future of how all medical records are stored and accessed? In Hadley’s study, the blockchain will guide how data flows between participants in the study and researchers. Women participating are able to revoke access to their data at any time.
We’re trying to teach computers to identify cancer in mammograms, and then teach radiologists what features to look for in an image,” Hadley said. “The holy grail of our work is to improve the art of medicine overall.”