Automated deep-neural-network surveillance of cranial images for acute neurologic events - NATURE MEDICINE
Titano, J. J. et al. Automated deep-neural-network surveillance of cranial images for acute neurologic events. Nat. Med. 24, 1337–1341 (2018).
Joseph Titano, M.D. and researchers from AISINAI used 37,236 head CT scans to train a deep neural network how to identify whether an image contained critical or non-critical findings. The platform was then tested in a blinded, randomized controlled trial in a simulated clinical environment where it triaged head CT scans based on severity. The computer software was tested for how quickly it could recognize and provide notification versus the time it took a radiologist to notice a disease. The average time for the computer algorithm to preprocess an image, run its inference method, and potentially raise an alarm was 150 times faster than physician identification. This study utilized “weakly supervised learning approaches,” which leveraged the research team’s expertise in natural language processing and the Mount Sinai Health System’s large clinical datasets.