Related Experiment Video
Updated: Oct 6, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.0K
Machine learning algorithms as new screening approach for patients with endometriosis
Sofiane Bendifallah1,2,3, Anne Puchar4,5, Stéphane Suisse6
1Department of Obstetrics and Reproductive Medicine, Hôpital Tenon, 4 rue de la Chine, 75020, Paris, France. sofiane.bendifallah@yahoo.fr.
Scientific Reports
|January 13, 2022
Summary
Machine learning algorithms show promise for diagnosing endometriosis using patient symptoms. This could lead to earlier detection and improved care for women with this chronic condition.
Area of Science:
- Reproductive Medicine
- Medical Informatics
- Artificial Intelligence in Healthcare
Background:
- Endometriosis is a complex, chronic condition affecting women of reproductive age with no routine diagnostic screening tests currently available.
- Despite research into biomarkers, genomics, and imaging, diagnostic laparoscopy remains the standard for endometriosis diagnosis, posing a barrier to timely care.
- There is a need for accessible and effective screening tools to aid in the early detection and management of endometriosis.
Purpose of the Study:
- To investigate the efficacy of machine learning algorithms (MLA) for the diagnosis and screening of endometriosis.
- To evaluate MLA performance using 16 key clinical and patient-reported symptom features.
- To explore the potential of MLA as a replacement for diagnostic laparoscopy in clinical practice.
Main Methods:
- Development and validation of machine learning algorithms utilizing 16 distinct clinical and patient-based symptom features.
- Performance evaluation of MLA using sensitivity, specificity, F1-score, and AUC metrics on training and validation datasets.
- Comparative analysis of MLA diagnostic capabilities against traditional diagnostic methods.
Main Results:
- MLA demonstrated high diagnostic performance, with sensitivity ranging from 0.82 to 1 and AUCs from 0.5 to 0.95 across training and validation sets.
- Specificity and F1-scores indicated robust performance, with AUCs consistently above 0.5, suggesting significant diagnostic potential.
- The developed MLA exhibited strong predictive capabilities for endometriosis diagnosis based on integrated symptom data.
Conclusions:
- Machine learning algorithms show significant potential as a non-invasive screening tool for endometriosis.
- MLA could serve as a valuable adjunct for general practitioners and gynecologists, facilitating earlier patient triage and diagnosis.
- Implementing MLA represents a paradigm shift, potentially replacing diagnostic laparoscopy and empowering patients in shared decision-making for endometriosis management.
