University of California San Francisco: AI for Imaging of Neurological Emergencies
This chapter aims to develop the artificial intelligence (AI) technology, including convolutional neural networks for deep learning; to extract quantitative biomarkers from head computed tomography (CT) scans in neurologic emergencies. It demonstrates that quantitative imaging biomarkers derived from the AI software are superior to subjective head CT grading schemes for patient outcome prediction in traumatic brain injury (TBI), hemorrhagic stroke, and aneurysmal subarachnoid hemorrhage. Imaging biomarkers would also not only streamline the painstaking and costly process of image interpretation by central expert readers but also provide more granular data for research into outcomes and therapies. Every 28 seconds, an American suffers a catastrophic neurologic emergency, most commonly stroke or TBI. Immediate diagnosis aided by rapid automated evaluation of head CT could greatly improve care in the emergency department; and even ambulances, intensive care units, and operating rooms are increasingly equipped with portable CT scanners.