Key points are not available for this paper at this time.
Background: Endometriosis is a chronic gynecological condition marked by ectopic endometrial-like tissue, leading to inflammation, pain, and infertility. Diagnosis is often delayed by up to 10 years. Identifying non-invasive biomarkers could facilitate earlier detection. MicroRNAs, known for their stability in biological fluids and role in disease processes, have emerged as potential diagnostic tools. This pilot study investigated whether serum miRNA profiling can differentiate endometriosis from other causes of chronic pelvic pain. Methods: Serum samples from 52 patients (36 with laparoscopically confirmed endometriosis and 16 controls) treated for chronic pelvic pain at a University Endometriosis Centre were analyzed. High-throughput miRNA sequencing was performed. Feature selection reduced 4285 miRNAs to the 20 most informative MiRNAs. Machine learning models, including logistic regression, decision tree, random forest, and support vector machine, were trained and evaluated. Results: Among the tested machine learning models, support vector machine achieved the best overall performance (accuracy 0.71, precision 0.80), while logistic regression and random forest showed the highest AUC values (0.84 and 0.81, respectively), indicating strong diagnostic potential of serum miRNA profiling. Conclusions: This study demonstrates the feasibility of using serum miRNA profiling combined with machine learning for the non-invasive classification of endometriosis. The identified miRNA signature shows strong diagnostic potential and could contribute to earlier and more accurate detection of the disease.
Kupec et al. (Mon,) studied this question.