Validation of a lab-free low-cost screening test for prevention of cervical cancer: automated visual evaluation

PROJECT SUMMARY/ABSTRACT Artificial intelligence (AI) has the potential to revolutionize medicine by improving productivity, reducing human error, and assisting with diagnosis and treatment. Image classification algorithms can be used to develop automated visual evaluation (AVE): a potential game-changer for cervical cancer prevention in low- and middle-income countries (LMICs). AVE technology reads digital photographs of a cervix to provide diagnosis and treatment recommendations in seconds. AVE is a true point of care test, low cost and does not require a laboratory. AVE could be used either for stand-alone primary screening, or to triage HPV-positive women. We will compare AVE to common screening methods in LMICs: visual inspection with acetic acid (VIA) and conventional cytology. Enhanced Visual Assessment (EVA) System by MobileODT is a cloud-connected mobile colposcope on a smartphone platform. It is FDA cleared and used in 42 countries. MobileODT is uniquely poised to integrate AVE into the EVA System. Our aim is to validate and commercialize AVE on the EVA platform. Phase I aims will adapt AVE to run on the EVA system using an optimal neural network architecture, running either directly on the phone or as a cloud- based service. Phase II is a prospective clinical trial of 10,000 patients recruited at ministry of health sites in El Salvador. All screen-positive patients, and 10% of negative patients, will undergo colposcopy with biopsy. Sensitivity of AVE as a primary screening test will be compared to cytology and to VIA. In HPV-positive women, AVE will be compared to VIA as a triage test.