Work

Risk Assessment System for Endometriosis in Medical Ultrasound Exams Using Machine Learning

Year: 2026
Type: article
Abstract: Endometriosis is a prevalent gynecological disease affecting millions of women worldwide, frequently causing chronic pelvic pain and infertility.Despite the widespread availability of pelvic ultrasound as a non-invasive diagnostic tool, its efficacy remains heavily dependent on examiner expertise, leading to notorious diagnostic delays.To address these limitations, this paper proposes a novel clinical decision support system based on serially organized machine learning models to automate and enhance endometriosis screening.Utilizing a real-world dataset of 7,020 medical exams (comprising 294,983 ultrasound images), it trained two complementary Multi-Layer Perceptron (MLP) expert models-one specialized in the positive class and the other in the negative class-integrated via an OR logic framework to strictly mitigate Type II errors (false negatives).Evaluated on a completely independent, out-of-sample test set of 6,080 exams, the proposed methodology achieved an overall Accuracy of 0.815 and an F1-score of 0.844.Crucially, the system reached a perfect Recall of 1.00, ensuring a complete absence of false negatives, which significantly outclasses existing deep learning baselines and traditional classifiers in the literature.These findings demonstrate that the proposed framework serves as a highly robust, low-and high-risk alert system that can optimize clinical monitoring strategies and enable timely therapeutic interventions.
Language: English
Open Access status: diamond
OpenAlex ID: W7203830017
FWCI:
Cites: 15
Cited by:
Related to: 9