A new machine learning algorithm developed by urologist Dr. Hsin-Hsiao (Scott) Wang and his team could revolutionize the way urologists predict vesicoureteral reflux (VUR) in infants with hydronephrosis. Hydronephrosis, a common congenital anomaly, is often identified during prenatal ultrasounds, but determining dilating VUR, a potential cause of hydronephrosis, remains a challenge.
Currently, urologists rely on the urinary tract dilation (UTD) classification system to evaluate hydronephrosis severity in infants based on ultrasound findings. However, this approach does not provide a specific diagnosis or predict outcomes. To address this gap, the team developed a machine learning model that analyzes early postnatal ultrasound features to predict the risk of dilating VUR.
The model, detailed in a recent study published in the Journal of Pediatric Urology, demonstrated strong predictive capabilities, with an area under curve of 0.81 out of 1.0. By leveraging distinct patient and imaging details, including demographics and UTD classification features, the algorithm can reliably determine which infants with hydronephrosis are more likely to have VUR and would benefit from further screening with a voiding cystourethrogram (VCUG).
Dr. Wang emphasizes that the model is designed to be user-friendly and easily integrated into routine clinical practice. The research team is currently expanding their dataset to develop a more comprehensive algorithm that could potentially predict the future course of hydronephrosis in each patient, shedding light on whether the condition will resolve on its own or require intervention.
This innovative approach holds promising implications for urologists, offering a potential ‘crystal ball’ to predict VUR in infants with hydronephrosis and optimize patient management strategies.
