Virginia Via Research Day Book 2026

Medical Student Research Clinical 14

THE VAGINAL HEALTH ASSESSMENT TEST FOR GYNECOLOGY PATIENTS WITH VAGINITIS: AN EXPERT READER VS TRAINED READER STUDY

Tayce Jaques, OMS-II; Haley Keene, OMS-II; Paushaly Sau, OMS-II; J. Mahaney, PhD Corresponding author: psau@vt.vcom.edu

VCOM-Virginia, Blacksburg, Virginia VCOM-Carolinas, Spartanburg, South Carolina Caza Health, LLC

Background: Vaginal infections are one of the most presented complaints by female patients. The use of artificial intelligence (AI) in this field has the potential to improve accuracy and subsequently provide more prompt treatment. Caza Health has developed the DayZ™ Vaginal Health Assessment Assay (VHA), which uses artificial intelligence (AI) coupled with immunofluorescence antibody cocktails plus automated scanning microscopy to determine the targets of interest, bacterial vaginosis, candida vaginitis, and trichomonas vaginosis, at the time of the office visit. Objective: The aim of this study is to evaluate the effectiveness of the DayZ™ VHA reader training program by comparing newly trained readers’ interpretations with expert-established ground truth (expert reader) to determine reader-to-reader variations. Methods: As part of the DayZTM development, a multi-reader study was conducted on a subset of samples, with the objective of assessing the ability of a newly trained reader to interpret results of the DayZ™ VHA test as compared to ground truth (expert reader). The effectiveness of the training was evaluated by

point-of-care instrument to provide physicians with improved vaginal biome data for the assessment and treatment of vaginitis. This study provides key information to enhance training effectiveness for end point users, including providing clearer definitions of clue cell characteristics and how to identify clue cells more accurately versus normal epithelial cells. Furthermore, this project will develop and improve training modules and methods to enhance easy start-up of the DayZTM VHA System when deployed in a clinician practice. IRB Approval: This study was approved by the VCOM Institutional Review Board (IRB), protocol 2020-017.

reader-to-reader variations and by discordant analysis with ground truth. Sixteen individual readers were trained one-on-one using the DayZ ™ System analysis software using the same exemplary image file. The readers then independently read image files in random order as their time allowed. Completed files by each reader were uploaded to Caza Health for analysis, which compared the results of the individual readers within the group for each target species and to the results to the ground truth (expert reader) who trains the AI algorithm. Results: At present, 200 (out of 400 planned) vaginal swab samples from symptomatic vaginosis participants have been analyzed, and algorithm training has achieved an accuracy of 86% for BV, 83% for CV and, 94% for TV compared to the results obtained by an expert human reader. As a result of the analysis, ways to enhance training effectiveness to better reader outcomes were identified including increased one-on-one training time, clearer definitions of clue cell characteristics, how to identify clue cells versus normal epithelial cells more accurately, and recognizing/rejecting non-specific staining, autofluorescence, and cellular debris. Conclusions: The DayZTM Vaginal Health Assessment System is in development as a near

Table of Contents

185 2026 Research Recognition Day