Understanding Deep Neural Network Predictions for Medical Imaging Applications
This work addresses the interpretability gap in medical imaging AI, which is crucial for real-world clinical adoption, though it appears incremental as it applies existing visualization techniques to new datasets.
The paper tackles the problem of uninterpretable decisions in deep neural networks for medical imaging by visualizing class activation mappings across four disease detection tasks, aiming to enhance model understanding for data scientists and assist doctors in decision-making.
Computer-aided detection has been a research area attracting great interest in the past decade. Machine learning algorithms have been utilized extensively for this application as they provide a valuable second opinion to the doctors. Despite several machine learning models being available for medical imaging applications, not many have been implemented in the real-world due to the uninterpretable nature of the decisions made by the network. In this paper, we investigate the results provided by deep neural networks for the detection of malaria, diabetic retinopathy, brain tumor, and tuberculosis in different imaging modalities. We visualize the class activation mappings for all the applications in order to enhance the understanding of these networks. This type of visualization, along with the corresponding network performance metrics, would aid the data science experts in better understanding of their models as well as assisting doctors in their decision-making process.