Biometric Collaborative Radiology Artificial Intelligence

Lung cancer is the leading cause of cancer death in the United States. with approximately 135.000 Americans dying of the disease each year. A 2017 report by the City of San Antonio Metropolitan Health District identified cancer as the 2 cause of death amongst Bexar County residents and predicted that by 2020 it would become the leading cause of death [1]. To mitigate its impact. the United State Preventative Task Force recommends annual radiologic screening with computed tomography (CT) in at risk individuals [2]. During this screening process. individuals undergo a chest CT which is interpreted by a physician radiologist. The radiologist identifies suspicious pulmonary lesions (nodules) and makes recommendations on further management which includes biopsy. continued surveillance. or more advanced methods of imaging. A national trial demonstrated that this radiologic surveillance method reduced mortality from lung cancer by approximately 16-20% [3]. The issue is that radiologist cancer detection is error prone. mitigation technology is inefficient and improperly tuned. and useful education approaches are lacking. Experts estimate that amongst positive radiological cases. 33% will have errors [4]. Each error represents a missed opportunity to detect malignancy. 
 In this proposal we propose a novel. comprehensive solution that addresses these problems through the synergism of human biometric science. technological advancement in artificial intelligence. and a rich collaborative network involving the UT San Antonio (UTSA). UT Health San Antonio (UT Health). and the Southwest Research Institute (SWRI). This method will not only deliver immediate benefits to the surrounding San Antonio community but it will also position San Antonio as an industrial leader in human biometric artificial intelligence. Radiologists will participate without intrusion as the technology can unobtrusively collect radiologist biometric data and be tuned to only intervene when a clear mistake is being made. Using a theme borrowed from autonomous vehicles. passive data collection by increasing numbers of radiologists will continuously improve accuracy at negligible cost which translates immediately into improved patient results. 
 Our core approach integrates unobtrusive eye-tracking technology. affixed underneath a radiologist monitor. which collects information on pupil size and screen gaze fixation. Based on known physiological principles these data points can be integrated with analyses of the image to ascertain if the radiologist is making a mistake. Unlike current state of the art implementations our approach avoids excessive false-positives (false alarms) which result in alarm fatigue and dismissal of error mitigation technology.