Researchers at chip company Nvidia, the Mayo Clinic, and the MGH & BWH Center for Clinical Data Science have built an algorithm which can creates various data sets for medical AI to study.
For artificial intelligence to work effectively, it relies on its set of information. A study accomplished under the collaboration of Nvidia, the Mayo Clinic, and MGH & BWH Center for Clinical Data Science has created a system that can produce diverse set of medical data. It created synthetic scans depicting abnormalities from existing MRIs of brain tumors using generative adversarial networks (GANs).
The system is a cutting-edge tool for training neural networks, but medical data will require more complication as “the data are usually imbalanced” said Hoo Chang Shin, a research scientist at Nvidia.
There is a great deal of abnormalities in medical cases that might be difficult for AI to diagnose as most of the data was programmed here. AI should be trained carefully before facing real-life situation especially in healthcare.