Evaluation of Vaginal Microbiome to Estimate Vaginal Health in Gynecology Patients Presenting With Vaginitis [ID: 1381290

Abstract

INTRODUCTION: Approximately 50% of vaginitis cases are misdiagnosed because of human error with the current diagnostic method of light microscopy wet mount. Untreated vaginitis may lead to complications of increased risk of sexually transmitted infections or preterm birth. Use of artificial intelligence (AI) and immunofluorescence scanning microscopy can aid in determining the health of the vaginal microbiome. METHODS: Institutional review board approval was obtained for the study. Participants who fit the inclusion and exclusion criteria were consented for participation. Patient demographic information was de-identified and recorded. Three swabs from each patient were collected: one for traditional wet-mount diagnosis, and two for DayZ Vaginal Health Assessment Assay analysis. The targets of interest recorded were clue cells, yeast pseudohyphae, and trichomonads. The evaluation compared wet-mount findings to the automated AI algorithm results. RESULTS: In preliminary data analysis of 58 patient samples, AI had an accuracy of 95% in finding trichomonads, 91% in candida, and 90% in clue cells. The AI detected trichomonads on one sample that was missed by the expert reader, detected pseudohyphae on three samples that were undetected by the expert reader, and detected clue cells in 17 samples that were undetected by the expert reader. These data are preliminary as the sample collection and data analysis are ongoing. CONCLUSION: The assay has shown superior ability to diagnose wet-prep findings compared to standard wet-mount evaluation. This new technology could provide a reliable source to determine a quick and accurate diagnosis for patients with symptomatic or asymptomatic vaginitis.