By: Dr. Eldad Elnekave, CMO.
When launching Zebra Medical Vision, we knew we had a talented team, immense imaging data and enough computational power to prove the potential of AI for the next generation of radiologists. The potential was endless, which was problematic.
The question was where to begin. Radiologists were contending with exponentially increasing images to interpret, greater resolution per image and an expanding repertoire of imaging bio-markers to recognize. We could harness our resources to help identify pertinent findings with greater precision and speed. Alternatively, we could begin by applying AI to find features considered less pertinent to the immediate reason for imaging a patient, but key to detection of diseases which would affect them in the future. If we could leverage the ubiquitous use of medical imaging to screen and risk-stratify populations for burdensome and common chronic conditions such as osteoporosis, cardiovascular disease and obstructive lung disease – we could impact the experience of healthcare beyond the radiologist. This path would be a longer one – creating the right algorithms would be just the first step. Next, we would need to prove their predictive value in a healthcare setting.
It was here that we decided to start (see blogs from May and November 2016), and so began our collaboration with Clalit Research Institute (CRI). Dr. Noa Dagan and Dr. Noam Barda led independent, 5 year retrospective cohort studies comparing CT based Zebra algorithms for prediction of osteoporotic fractures and cardiovascular events to the gold standard contemporary predictive models of the FRAX fracture assessment tool and the American Heart Association, respectively. The studies, which included 48,227 and 14,135 subjects respectively, demonstrate the power of AI used in routine medical imaging to re-conceive the scope of screening and risk-stratification for major chronic illness. The work will be presented first at the 2018 Radiological Society of North America (RSNA) on November 27th and 28th.
Almost five years ago we decided to invest in a path which has the potential to change not only the experience of the radiologist but even that of a generation at risk of osteoporosis. And we didn’t stop there, but rather used each algorithm developed as a platform to develop the next, allowing us to go forward with AI in breast cancer detection , automatic Chest X-ray analysis and detection of acute intracranial bleeds . What’s next?