AI summaryⓘ
The authors developed a new way to understand how deep neural networks make decisions in medical image classification by looking at how small changes in the input affect the model's predictions. They use a mathematical tool called the input-dependent Fisher Information Matrix (iFIM) to measure this sensitivity, focusing on key directions that strongly influence the model's output. Unlike common methods that highlight important pixels, their approach gives insight into the model's local decision process itself. They tested this approach on various medical tasks and found that the most sensitive directions identified by iFIM closely relate to changes in prediction confidence and accuracy. This method offers a more direct and principled way to analyze model behavior alongside existing interpretability techniques.
Deep Neural NetworksMedical Image ClassificationInterpretabilityFisher Information MatrixInput SensitivityPredictive DistributionEigenspectrumGram MatrixLocal Sensitivity AnalysisPost-hoc Interpretation
Abstract
Deep neural networks have achieved strong performance in medical image classification, but often work like black-box. Commonly used post-hoc interpretation methods often provide heuristic visualizations whose relationship to the classifier's predictive distribution is indirect. This work introduces a local sensitivity analysis framework based on the input-dependent Fisher Information Matrix (iFIM) of a trained classifier. The iFIM characterizes how the classifier's predictive distribution changes under infinitesimal perturbations of the input image. By using a Gram-matrix formulation, the nonzero eigenspectrum of the iFIM can be recovered without explicitly forming the full image-dimensional Fisher matrix. The leading iFIM eigenspace is then used to project an input image into a high local-sensitivity component and its orthogonal component. These components provide a model-intrinsic description of local predictive sensitivity, rather than a conventional pixel-wise attribution heatmap or a causal segmentation of task-relevant anatomy. The framework is evaluated on controlled and clinical medical image classification tasks using multiple classifier architectures. Perturbation-based experiments show that high-sensitivity iFIM components are more strongly coupled to changes in predictive confidence and classification performance than lower-sensitivity complementary components. The results support the iFIM framework as a principled tool for analyzing local decision sensitivity and for complementing existing attribution-based interpretability methods in medical imaging.