Using Natural Language Processing to Support Early Detection of Endometriosis

Content uploaded by Keira MacDonald Author content All content in this area was uploaded by Keira MacDonald on Jan 14, 2026 Content may be subject to copyright. ZU064-05-FPR output 14 January 2026 14:2 1 Using Natural Language Processing to Support Early Detection of Endometriosis Keira Sian MacDonald1,2 1University of Western Ontario, London, ON, Canada 2University of Warwick, CV4 7AL, UK (e-mail: [email protected]) Abstract Endometriosis is a chronic inflammatory condition that affects approximately 190 million women who are assigned female at birth. Endometriosis lesions cause severe, often debilitating pain. De- spite its prevalence and impact, endometriosis remains one of the most misdiagnosed conditions in women’s health. Endometriosis diagnoses are often delayed several years due to the subjective nature of its symptoms. As a result, clinically relevant information exists primarily in patient nar- ratives. This study explores natural language processing (NLP) techniques to extract meaningful diagnostic signals from patient narratives. The aim of this study is to support earlier identification of individuals at risk of endometriosis. Anonymous patient narratives are analysed using NLP to parse common symptom descriptions and pain descriptors, which is used to train supervised machine learning models to differentiate narratives associated with diagnosed endometriosis and control cases. Model interpretability techniques are applied to identify language patterns associated with risk. The expected outcome is that NLP-based models will identify language patterns characteristic of endometriosis, including cyclic pain descriptions and evolving symptom predictions over time. These models are expected to outperform baseline heuristic approaches based on isolated symptom indi- cators. This suggests that patient narratives contain underutilized diagnostic signals relevant to early clinical consideration. This study highlights the feasibility of using NLP-driven analysis of patient- reported data as a scalable and non-invasive support tool for earlier identification of endometriosis. It demonstrates the potential of computational methods to combat diagnostic delays in conditions that may be widely experienced but underrepresented, providing a possible framework to support other underdiagnosed diseases. 1 References [1] Bontempo AC, Mikesell L. 2020. Patient Perceptions of Misdiagnosis of Endometrio- sis: Results from an Online National Survey. Diagnosis (Berl). May 26;7(2):97-106. doi: 10.1515/dx-2019-0020. PMID: 32007945. [2] Horne A. W., Missmer S. A. 2022. Pathophysiology, Diagnosis, and Management of Endometriosis. BMJ: British Medical Journal, 379, 1–19. https://www.jstor.org/stable/27424071. [3] Ranney, B. 1952. Endometriosis. The American Journal of Nursing, 52(12), 1465–1467. https://doi.org/10.2307/3459462. ZU064-05-FPR output 14 January 2026 14:2