(Poster 96) Classifying Pelvic Floor Muscle Strength Using Machine Learning on ESPP

Objectives: To develop and validate a machine learning model for classifying pelvic floor muscle (PFM) strength using non-invasive extracorporeal surface perineal pressure (ESPP) measurements.

Methods: A total of 23 female participants underwent ESPP measurements using a 10×10 pressure sensor array. PFM strength was assessed using transabdominal ultrasound to measure bladder base displacement. K-means clustering was used to establish cut-off values for weak and strong PFM groups. Six machine learning algorithms, including random forest, support vector machine, and AdaBoost, were trained and tested on ESPP data to classify PFM strength. Model performance was evaluated using area under the curve (AUC), accuracy, and F1 score. The most predictive ESPP features were identified using feature mutation importance and Shapley additive explanations (SHAP).

Results: The random forest algorithm demonstrated the highest performance in classifying weak versus strong PFM strength, with an AUC of 0.920, accuracy of 0.900, and F1 score of 0.905. Feature importance analysis indicated that directional contraction vectors, particularly the middle, lateral, and lower front vectors, were more predictive of PFM strength than overall contraction pressure.

Conclusions: Machine learning applied to ESPP measurements offers a promising non-invasive approach for assessing PFM function. This method provides an objective and comprehensive evaluation of PFM strength, potentially improving the diagnosis and treatment of pelvic floor disorders. By leveraging machine learning techniques, clinicians may gain deeper insights into PFM biomechanics, allowing for more personalized treatment strategies and rehabilitation approaches for pelvic floor dysfunction.