Usability analysis of integrated and improvised smartphone-based cervical imaging systems

Cervical cancer remains a leading cause of cancer mortality in low- and middle-income countries, where screening programs often rely on visual inspection with acetic acid (VIA) despite variable accuracy and reproducibility. While Automated Visual Evaluation (AVE) systems that pair cervical imaging with artificial intelligence may improve consistency, their effectiveness depends on reliable image capture under real-world constraints. Here, we conducted a comparative usability evaluation of two smartphone-based AVE-assisted cervical imaging systems used in the OPTICS trial in El Salvador: an integrated mobile colposcope (MobileODT EVA) and an improvised system using a mounted Samsung S20 smartphone with auxiliary lighting. Paired data from 3,649 participants were analyzed using two usability surrogates: exam duration derived from CervManager™ timestamps and image sharpness assessed via expert labeling of blur. Among 3,408 eligible records, the EVA system produced a lower proportion of exams with at least one sharp image than the S20 (85.9% vs 98.8%) and required more images per exam (median 4 vs 3). EVA durations were more tightly clustered with fewer extreme delays, while S20 durations showed greater variability and a right-skew driven by one provider early in the study. Blur and exam duration patterns, including improvements after retraining, indicate that usability challenges are strongly influenced by training and onboarding rather than device limitations.