From Single-Visit to Multi-Visit Image-Based Models: Single-Visit Models are Enough to Predict Obstructive Hydronephrosis

TL;DR

This study evaluates whether incorporating multiple ultrasound visits improves prediction of obstructive hydronephrosis, finding that using only the latest ultrasound is generally sufficient, simplifying clinical decision-making.

Contribution

The paper demonstrates that multi-visit ultrasound data offers minimal benefit over single-visit data for predicting obstructive hydronephrosis, challenging the need for complex multi-visit models.

Findings

Incorporating past visits provides limited predictive improvement.

Single-visit ultrasound data is nearly as effective as multi-visit data.

Simple models using only the latest ultrasound suffice for risk prediction.

Abstract

Previous work has shown the potential of deep learning to predict renal obstruction using kidney ultrasound images. However, these image-based classifiers have been trained with the goal of single-visit inference in mind. We compare methods from video action recognition (i.e. convolutional pooling, LSTM, TSM) to adapt single-visit convolutional models to handle multiple visit inference. We demonstrate that incorporating images from a patient's past hospital visits provides only a small benefit for the prediction of obstructive hydronephrosis. Therefore, inclusion of prior ultrasounds is beneficial, but prediction based on the latest ultrasound is sufficient for patient risk stratification.