An Explainable AI-Based Decision Support Tool to Predict Preterm Birth

Ilias Kyparissidis-Kokkinidis, Emmanouil S. Rigas, Evangelos Logaras, Ioannis Tsakiridis, Themistoklis Dagklis, Paris Lagakis, Antonis Billis, Panagiotis D. Bamidis

Keywords: Artificial Intelligence, AI, Machine Learning, ML, Explainable Artificial Intelligence, XAI, Decision Support Systems, Clinical, Premature Birth, Obstetrics and Gynecology

TL;DR: An Explainable AI-Based Decision Support Tool to Predict Preterm Birth

Abstract: Preterm Birth (PTB), characterized as birth occurring prior to 37 weeks of gestation, presents a notable clinical challenge. In this work, we aim to assist the decision making process of the obstetricians by proposing an AI-based clinical decision support system. Specifically, we propose a Machine Learning (ML)-based model to efficiently predict PTB using an assortment of relevant features including social demographics, medical history, along with laboratory and obstetric examination results. This model was trained and validated using a dataset consisting of 873 women from a major Greek hospital. Moreover, we implemented an explainability feature using SHapley Additive exPlanations (SHAP) to enhance the clinical interpretation of the results. Additionally, we developed a web application that encompasses both the predictive model and the explainability feature in an easy-to-use user interface. The predictive model has shown strong performance in internal validation, achieving an accuracy of 94% and a recall of 97%. In external validation, where the tool was used by clinicians for 100 pregnant women, it achieved an accuracy of 89% and a recall of 94.3%. Finally, the web application was well accepted by the clinicians.

Track: 4. AI-based clinical decision support systems

Registration Id: KQNBLTHX5ZF

Submission Number: 161