Domain Adaptation for Deviating Acquisition Protocols in CNN-based Lesion Classification on Diffusion-Weighted MR Images
This addresses the challenge of deploying deep learning models in heterogeneous clinical settings for breast cancer classification, though it is incremental as it builds on existing domain adaptation techniques.
The paper tackled the problem of CNN-based lesion classification on diffusion-weighted MR images being dependent on specific input channels used during training, which limits large-scale application due to protocol deviations between clinical sites. They proposed a model-based domain adaptation method that restores training inputs from altered channels during deployment, resulting in a significant increase in classification performance and superiority over implicit adaptation methods.
End-to-end deep learning improves breast cancer classification on diffusion-weighted MR images (DWI) using a convolutional neural network (CNN) architecture. A limitation of CNN as opposed to previous model-based approaches is the dependence on specific DWI input channels used during training. However, in the context of large-scale application, methods agnostic towards heterogeneous inputs are desirable, due to the high deviation of scanning protocols between clinical sites. We propose model-based domain adaptation to overcome input dependencies and avoid re-training of networks at clinical sites by restoring training inputs from altered input channels given during deployment. We demonstrate the method's significant increase in classification performance and superiority over implicit domain adaptation provided by training-schemes operating on model-parameters instead of raw DWI images.