Explanation-guided framework for reconstructing missing multimodal clinical features while preserving information relevant to pancreatic cancer survival prediction.
Training-free method that uses gradient-driven smooth contours for compact visual attribution.
Fourier Shape Representation is a repository and research project focused on using Fourier-based techniques to represent and analyze shapes in images. The project implements …
A graph neural network and gaze-guided transformer framework to predict radiological diagnostic errors from eye-tracking data.
Extremal Contours is an algorithmic explainable AI method for extracting compact, smooth, and interpretable visual explanations from deep vision models. The method represents …
A novel shape-aware loss function for anatomical segmentation that leverages PCA to enforce realistic segment morphology.