EP24.13: Advancing endometriosis mapping: quantifying sliding sign in transvaginal sonography to identify Pouch of Douglas obliteration

This project aims to enhance the reliability of endometriosis mapping through Transvaginal Sonography (TVS) by developing an image processing technique that quantifies the sliding sign between the uterus and adjacent tissues with a special focus on POD. The primary goal is to provide a more objective and consistent method for detecting endometriosis based on tissue interaction observed in TVS. Additionally, we introduce the concept of incorporating tissue elasticity into the analysis using elastic registration techniques to improve the accuracy of solving equations related to the relative displacement between two tissues (the parameter will be solved in tissue interaction). We developed a specialised image processing algorithm using Python and OpenCV. This involved extracting frames from TVS videos, manually selecting regions of interest, and applying motion tracking algorithms. We calculated the relative displacement of tissues over time, considering tissue elasticity, and created visual representations with numerical maps indicating relative displacement. To validate, we rigorously assessed its effectiveness with the opinion of two expert radiologists in the field of gynecological imaging. Normal cases displayed typical tissue movement patterns, while pathological cases showed restricted or altered movements, indicating adhesion presence. Results were visualised through relative displacement plots, offering clear and interpretable data. We tested on 10 normal (normal POD) and 5 pathological cases (obliterated POD), ensuring high accuracy in differentiating observed states during TVS (with 94% diagnostic accuracy). The developed image processing technique is a promising tool for radiologists, potentially improving endometriosis mapping accuracy. It could benefit less experienced radiologists, offering a quantitative basis for assessing the sliding sign in TVS exams and facilitating information sharing between radiologists and clinicians.