Diagnostic Accuracy of an Order Parameter-Based Image Analysis Method From Cervical Images [ID 1504]

INTRODUCTION: Cervical histopathology is the gold standard for diagnosing cervical cancer and pre-cancer but poses significant barriers in low- and middle-income countries, highlighting the need for accessible alternatives. Through a collaboration with physical scientists and clinicians, our study aims to validate the accuracy of a modified Bragg Williams order parameter analysis for diagnosing cervical dysplasia using digital colposcopy images. METHODS: A retrospective cohort study was conducted using the IARC database analyzing 172 cervical images with the modified Bragg Williams order parameter image analysis method. One colposcopy image per patient was selected and blindly interpreted with S2 values ranging from 0 (complete disorder) to 1 (perfect order). Images were categorized by histologic diagnoses: normal, CIN 1, 2, 3, or cancer. An additional 137 images were blindly analyzed to calculate sensitivity, specificity, and area under the curve. RESULTS: Higher grades of dysplasia corresponded with lower degrees of order. S2 ranges were 0.685–0.8613 for 80 benign images, 0.5842–0.6329 for 18 CIN 1 images, 0.5124–0.5481 for 20 CIN 2 images, and 0.3390–0.5022 for 29 CIN 3 images. An additional 137 images were blindly analyzed, and the ranges accurately diagnosed normal CIN 1, 2, 3, and cervical cancer with perfect accuracy. CONCLUSIONS/IMPLICATIONS: The study assigned specific degrees of order to various squamous cell abnormalities of the cervix, demonstrating the method's accuracy. Future studies will use diverse clinical settings and colposcopy equipment to further validate this technology.