Confounding variables can degrade generalization performance of radiological deep learning models

Zech, John R.; Badgeley, Marcus A.; Liu, Manway; Costa, Anthony B.; Titano, Joseph J.; Oermann, Eric K.

doi:10.48550/arxiv.1807.00431

Confounding variables can degrade generalization performance of radiological deep learning models

Early results in using convolutional neural networks (CNNs) on x-rays to diagnose disease have been promising, but it has not yet been shown that models trained on x-rays from one hospital or one group of hospitals will work equally well at different hospitals. Before these tools are used for com…