A team from the Massachusetts General Hospital was among the researchers talking about how they’re using AI at GTC DC earlier this year.
The Mass General researchers joined colleagues from across the healthcare industry to help tell the story of how deep-learning – which is already used by hundreds of millions of people on smartphones – can improve health care.
Mass General became the first medical institute in the world — and among the first five research institutions of any kind — to receive an NVIDIA DGX-1. We delivered the supercomputer at Mass General’s historic Ether Dome, where the first public demonstration of surgery using anesthetic took place in 1846.
Mass General’s Clinical Data Science Center joins other early DGX-1 users, including the Open AI Institute, the Stanford Artificial Intelligence Laboratory, benevolent.ai, SAP and the Berkeley Artificial Intelligence Research Lab.
The center is already using GPUs to make significant medical advances. Researchers are testing an automated bone-age analyzer they’ve created that speeds diagnosis of children’s growth problems and is nearly as accurate as human radiologists (see “Deep Learning Speeds Diagnosis of Kids’ Growth Problems”).
More is coming. The Clinical Data Science Center is using AI and deep learning to advance healthcare, beginning with radiology, pathology and genetics. The center will research, test and implement new ways to improve the detection, diagnosis, treatment and management of diseases by training a deep neural network using Mass General’s vast stores of phenotypic, genetics and imaging data. The hospital has a database containing 10 billion medical images.
“The intent is to be able to explore the integration of man and machine at this point of clinical care, taking some of the data historically and using that data to actually create information in the machine so that we can see into the future what’s happening with patients before the human has the idea that there are changes taking place,” said Dr. Keith Dreyer, vice chairman and associate professor of radiology at Mass General and Harvard Medical School and executive director of the Mass General Clinical Data Science Center.
Radiology and Medical Imaging
DGX-1 also promises to help accelerate the adoption of AI in fields where machine learning techniques have already made a difference, such as radiology and medical imaging.
“The importance of machine learning and machine learning for radiology is unquestioned,” said Dr. James Brink, head of radiology at Mass General and chair of the American College of Radiologists. “I think there’s an enormous amount of opportunity for us to improve the efficiency of our work and the accuracy of our work through automation and semi-automation.”
Work with Patients
Longer term, deep learning also promises to help deliver better care for today’s patients by letting doctors better use the flood of medical research and patient data being produced by Mass General and other medical centers.
“I see deep learning and other machine learning techniques that could help us on a day-to-day basis make the process more efficient and in essence even more accurate,” said Dr. Long Li, assistant in pathology at Mass General and an assistant professor of pathology at Harvard Medical School.
Sounds like just what the doctor ordered.