U-Noise: Learnable Noise Masks for Interpretable Image Segmentation
This work addresses the need for interpretability in deep neural networks used in critical applications like medical imaging, though it appears incremental as it builds on existing explainability methods.
The paper tackles the problem of interpreting image segmentation models by introducing a method that learns noise masks to identify image regions where noise does not affect model performance, applied to pancreas segmentation in CT scans with qualitative comparisons to existing techniques and a quantitative evaluation based on downstream performance.
Deep Neural Networks (DNNs) are widely used for decision making in a myriad of critical applications, ranging from medical to societal and even judicial. Given the importance of these decisions, it is crucial for us to be able to interpret these models. We introduce a new method for interpreting image segmentation models by learning regions of images in which noise can be applied without hindering downstream model performance. We apply this method to segmentation of the pancreas in CT scans, and qualitatively compare the quality of the method to existing explainability techniques, such as Grad-CAM and occlusion sensitivity. Additionally we show that, unlike other methods, our interpretability model can be quantitatively evaluated based on the downstream performance over obscured images.