Interpretable Deep Learning for Medical Imaging

Interpretable Deep Learning for Medical Imaging related Clinical Decision Support

 Project Team: Diane Gromala (Interactive Arts and Technology, SFU), Ghassan Hamarneh (Computing Science, SFU), Sheelagh Carpendale (Computing Science, SFU), Weina Jin (Interactive Arts and Technology, SFU).

Although Big Data and Deep Learning (DL) in medicine are advancing rapidly, their implementation into patient-care settings has not yet become widespread. One pivotal impediment is the black-box nature of DL models. Since physicians typically do not have an explanation for the model’s output, they are undertaking high risks if the models are predicting erroneously.

This project will develop and evaluate an interpretable DL system on medical imaging data that is explainable to doctors for their clinical decision support. We combine DL with visualization and interaction techniques to make it more interpretable. We then deploy the system at our partner hospital for identifying neuromuscular disorders. We will evaluate the resulting system with both DL practitioners and physicians. Both large, publicly available medical imaging datasets and the private MRI dataset from our clinical collaborators will be utilized. This project is interdisciplinary research of big data, information visualization, human-computer interaction and medicine.