Utilization of Machine Learning for the Detection of Detrusor Overactivity on Urodynamics
Kevin T. Hobbs, MD1, Nathaniel Choe, BSE2, Leonid I. Aksenov, MD1, Wilkins Aquino, PhD2, Jonathan C. Routh, MD, MPH1, James A. Hokanson, PhD3.
1Duke University Medical Center, Durham, NC, USA, 2Duke University, Pratt School of Engineering, Durham, NC, USA, 3Medical College of Wisconsin, Department of Biomedical Engineering, Milwaukee, WI, USA.
Background: Voiding dysfunction is a significant cause of morbidity and quality-of-life impairment in children with neurogenic bladder, particularly those with spina bifida. Urodynamic studies (UDS) can help identify those patients at risk for harmful sequelae, such as renal damage or recurrent urinary tract infections, and guide clinical decision making. Unfortunately, a limitation of UDS is the subjective variability in the interpretation of urodynamic findings. Machine learning, a subset of artificial intelligence that uses dynamic algorithms to identify patterns and improve accuracy as more data is analyzed, has the potential to help standardize UDS interpretation. Our objective for in this study was to assess the accuracy of various machine learning models at detecting the presence or absence of detrusor overactivity (DO).
Methods: Patients who underwent UDS at a single pediatric urology clinic between 5/2012 and 9/2020 were included. UDS files were analyzed in both time and frequency domains, varying inclusion of vesical, abdominal, and detrusor pressure channels. A machine learning pipeline was constructed using data windowing, dimensionality reduction, and support vector machines. Models were designed to detect clinician identified detrusor overactivity.
Results: Data were extracted from 805 UDS testing files from 546 unique patients. The generated models achieved good performance metrics in detecting DO agreement with the clinician, in both time- and frequency-based approaches. Incorporation of multiple channels and data windowing improved performance (Figure). The time-based model with all 3 channels had the highest AUC (91.9 ± 1.3%; sensitivity: 84.2 ± 3.8%; specificity: 86.4 ± 1.3%). The 3-channel frequency-based model had the highest specificity (AUC: 90.5 ± 1.9%; sensitivity: 68.3 ± 5.3%; specificity: 92.9 ± 1.1%).
Conclusions: We developed a promising proof-of-concept machine learning pipeline that identifies DO in UDS. Machine-learning-based predictive modeling algorithms can be used to standardize UDS interpretation, augment shared decision-making, and potentially improve the life of pediatric patients and reduce costs to the healthcare system.
Back to 2021 Abstracts