PhD Scientific Days 2025

Health Sciences III.

Digital Symptom Tracking and Machine Learning for Endometriosis Prediction: Insights from the Lucy App

Előadó neve

Balogh Dora Bianka, PhD

Neptun code

IY6NRF

Előadó munkahelye

Semmelweis University

Előadó telefonszáma

+36307895690

Előadó e-mail címe

dorabiankabalogh@gmail.com

Az előadás címe

Digital Symptom Tracking and Machine Learning for Endometriosis Prediction: Insights from the Lucy App

Szerző(k) neve és munkahelye

Dora Bianka Balogh¹, Dmitrijs Bļizņuks², Reka Brubel¹, Gernot Hudelist³, Nandor Acs¹, Attila Bokor¹

1: Semmelweis University
2: Riga Technical University
3: Hospital St. John of God

Bemutatás módja

Szóbeli

Szekció

Health Sciences III.

Language of the presentation

English

Preferred session

Health Sciences

Összefoglaló szövege

Introduction
Limited funding and research have stalled medical innovation in endometriosis, contributing to diagnostic delays of 4-11 years. Mobile healthcare tools can improve chronic disease management, symptom assessment, and prediction, but their application in endometriosis remains scarce. This study tests predictive machine learning (ML) models on Lucy app self-reported data.

Methods
Lucy app collects self-reported data from individuals with endometriosis and healthy controls. After filtering, the preliminary analysis was performed on 520,000 user records, utilizing correlation methods and visual tools to explore associations between symptoms and endometriosis. An XGBoost and Random Forest methods were implemented to identify key symptom patterns and predict endometriosis, training on the subset of 4812 users (n = 1212 control, n = 3600 endometriosis). This study is registered at ClinicalTrials.gov (Identifier: NCT06147687).

Results
The first step involved data filtering to address issues like nonrealistic dates and inconsistencies typical in real-world datasets. Strong associations between known endometriosis-associated symptoms—including pelvic pain, pelvic cramps, dysmenorrhea, and lower back pain—were shown by records from patients diagnosed with endometriosis but not yet treated. Our Random Forest model achieved an overall accuracy of 0.84, with an F1-score of 0.78 for endometriosis and 0.87 for negative cases, indicating effective discrimination between endometriosis and control cases. These metrics were obtained as mean values from 5-fold cross-validation, supporting the robustness of the model. Using Feature Importance analysis, dysmenorrhea and pelvic pain were identified as the most impactful symptoms on classification decisions.

Conclusion
Our preliminary analysis demonstrated that real-world, self-reported data are reliable and consistent with known endometriosis-associated symptoms, suggesting that the use of mobile apps like Lucy for endometriosis monitoring is a promising strategy. These findings may pave the way to transforming disease management and early detection of endometriosis.

Funding
This study was funded by the Horizon 2020 Research and Innovation Programme (grant number 101017562) and by the New National Excellence Program of the Ministry for Culture and Innovation (grant number EKÖP-2024-71).

University

Semmelweis University

Supervisor

Attila Bokor, MD, PhD

Publication of my abstract

I do not give consent to the publication of my abstract on the website of the congress.

phd.section.field

after finishing doctoral studies with absolutorium (PhD)

Kind

Szabad

Status

elfogadva

Accepted presentation method

szóbeli

Előadás fájl jóváhagyás

nem hagyta jóvá

Előadó

4068

Start

10:15

End

10:30