Subscribe to our newsletter

The University of California, San Francisco and GE Healthcare are studying how artificial intelligence and machine learning can help doctors and caregivers make faster and smarter clinical decisions. Together, they will be developing deep learning algorithms aimed at delivering information to clinicians faster.[1]

Versions of AI such as those found in Apple’s Siri or Microsoft’s advanced image-recognition system have begun to prove the technology’s capability, “but in healthcare, there has not been nearly as much progress,” says Dr. Michael Blum, associate vice chancellor for informatics and a cardiologist at UCSF. “In medical school, physicians learn to use a stethoscope and to read X-rays to help identify what’s happening inside a patient’s body. Now, we will augment those century-old tools with contemporary technologies including artificial intelligence and machine learning.”

An image of the lungs adjusted for viewing in virtual reality goggles. Image and GIF credit: GE Healthcare/GE Reports

The UCSF and GE Healthcare team will first develop and validate AI algorithms using thousands of anonymized and annotated chest X-rays, many acquired using GE Healthcare equipment. Once the solution is deemed safe and effective, it can then be deployed worldwide on the GE Health Cloud and smart GE Healthcare imaging machines, and will have the ability to analyze large volumes of X-rays for critical abnormalities, such as a collapsed lung or an inappropriately placed feeding tube.

The technology in development aims to make clinical care teams more efficient and to help radiologists more intelligently prioritize their work by pushing cases that the AI algorithms identify as critical to the top of their work list. The long-term goal is to reduce the time it takes to treat patients in acute situations and improve patient outcomes.

Dr. Blum says that without the support of such algorithms, the radiologist’s time is not always effectively utilized. A radiologist, for example, might look at dozens of normal or unchanged chest X-rays before reviewing an exam with a time-sensitive imaging finding. The science behind deep learning enables a radiologist to provide the system with valuable feedback by confirming or rejecting the software’s selection, and continuously feeding it with new imaging data that constantly improves the accuracy of the algorithm.

“In medical school, physicians learn to use a stethoscope and to read X-rays to help identify what’s happening inside a patient’s body,” says UCSF’s Dr. Blum. “Now, we will augment those century-old tools with contemporary technologies including artificial intelligence and machine learning.”

GE Healthcare’s AI development roadmap aims to develop a library of algorithms for all diagnostic imaging methods, helping to improve diagnostic accuracy and patient outcomes as well as clinical workflows and productivity.

Dr. Blum says the first algorithms will be developed and tested over the coming six months and will focus on supporting clinicians in their daily practice. “They won’t be making a diagnosis or recommending a treatment initially, but we hope to develop those more sophisticated algorithms as the collaboration progresses. It’s easy to imagine that eventually we will develop algorithms that are numerically as good as the doctors [at making a diagnosis], but there will always be the need for experienced physicians in the complex, emotional undertaking of providing healthcare.”

This statement especially resonates in some global healthcare markets, including emerging markets, where there is a shortage of radiologists and radiology specialists. The future algorithms have the potential to address a lack of clinical resources and ensure providers around the world can access new knowledge and insights delivered through deep learning.

GE Healthcare and UCSF’s collaboration brings together two teams with a storied history in the field of diagnostic imaging. GE Healthcare invented the X-ray in 1895, and UCSF opened one of the first dedicated X-ray facilities in 1912 to instruct all medical students in radiology. Today, their partnership is helping shape the future of patient care.

[1] Technology in development that represents ongoing research and development efforts. These technologies are not products and may never become products. Not for sale. Not cleared or approved by the U.S. FDA or any other global regulator for commercial availability