Detection of obstetric anal sphincter injuries using machine learning-assisted impedance spectroscopy: a prospective, comparative, multicentre clinical study

Abstract

AbstractObjectiveTo evaluate the clinical performance and safety of the ONIRY system for obstetric anal sphincter injuries (OASI) detection versus three-dimensional endoanal ultrasound (EAUS).DesignA prospective, comparative, multicentre, international study. Setting: Poland, Czechia, Slovakia, and Spain.Population152 women between the first moments up to 8 weeks after vaginal delivery.MethodsParticipants underwent EAUS and were allocated to groups based on OASIS classification: A (no perineal tear), B (1st or 2nd degree tear), or C (3rd or 4th degree, anal sphincters affected). Electric impedance was measured in the anal canal using the ONIRY system. The primary endpoint was the diagnostic outcome of impedance spectroscopy versus EAUS. Adverse events were collected. Part II involved in silico modelling and 10-time 10-fold cross-validation for automated analysis.Main Outcome MeasuresAccuracy, sensitivity, and specificity.Results30 women were allocated to group A, 61 to group B, and 61 to group C. The diagnostic outcome was determined for 147 participants. The accuracy, sensitivity, and specificity of the ML-assisted impedance spectroscopy were 87.0 ± 0.5%, 90.6 ± 2.0%, and 84.6 ± 1.9%, respectively, compared with EAUS. After data cleaning, the performance metrics of the proposed final ML model for ONIRY were: 90.0 ± 0.4%, 90.0 ± 1.2%, and 90.0 ± 0.7%, respectively. No adverse device effects or deficiencies were observed.ConclusionThe ML-assisted impedance spectroscopy appears to be a high-performance and safe method for rapid OASI detection, complementing digital rectal examination in obstetric settings.FundingThe study was financed by the Polish National Centre for Research and Development (POIR.01.01.01-00-0726/18).