Aggregated Attributions for Explanatory Analysis of 3D Segmentation Models

TL;DR

This paper introduces Agg^2Exp, a method for aggregating voxel attributions to improve explainability of 3D segmentation models, especially in medical imaging, by providing a global view of segment importance and demonstrating gradient-based attribution's fidelity.

Contribution

The paper presents Agg^2Exp, a novel aggregation technique for voxel attributions that enhances global interpretability of 3D segmentation models, surpassing classical local explanation methods.

Findings

Gradient-based attributions are more faithful than perturbation-based explanations.

Agg^2Exp provides a comprehensive global view of segment importance.

Application to medical imaging demonstrates its practical utility.

Abstract

Analysis of 3D segmentation models, especially in the context of medical imaging, is often limited to segmentation performance metrics that overlook the crucial aspect of explainability and bias. Currently, effectively explaining these models with saliency maps is challenging due to the high dimensions of input images multiplied by the ever-growing number of segmented class labels. To this end, we introduce Agg^2Exp, a methodology for aggregating fine-grained voxel attributions of the segmentation model's predictions. Unlike classical explanation methods that primarily focus on the local feature attribution, Agg^2Exp enables a more comprehensive global view on the importance of predicted segments in 3D images. Our benchmarking experiments show that gradient-based voxel attributions are more faithful to the model's predictions than perturbation-based explanations. As a concrete use-case, we apply Agg^2Exp to discover knowledge acquired by the Swin UNEt TRansformer model trained on the TotalSegmentator v2 dataset for segmenting anatomical structures in computed tomography medical images. Agg^2Exp facilitates the explanatory analysis of large segmentation models beyond their predictive performance. The source code is publicly available at https://github.com/mi2datalab/agg2exp.