Impact of image capture assistance tools and devices on the exam adequacy rate of automated visual evaluation of cervical images (Conference Presentation)

Most cervical cancer deaths occur in low-resource settings. Automated visual evaluation (AVE) is a screening method based on deep learning and imaging. AVE uses quality classifiers during image capture. Low-quality images are excluded from diagnostic evaluation. The relationship between these different classifiers and imaging devices is poorly understood. An AVE application was installed on a mobile colposcope and a standalone phone. Six printed image phantoms were imaged. The effect of AVE quality modules on the satisfactory AVE exam rate (those with minimum one qualified image). Three of 6 standalone phone exams were found unsatisfactory, yet all 6 mobile colposcope AVE exams were satisfactory.