Adaptive Group-Based Counterfactual Explanations for Time-Series Rehabilitation Data

2026-07-02

AI Summary

This study addresses the challenge that existing counterfactual explanation methods struggle to generate coherent, biomechanically plausible explanations aligned with clinicians’ abstract understanding of muscle groups and joint segments in multichannel time-series rehabilitation data. To overcome this limitation, the authors propose a two-stage framework: first employing Shapley-Adaptive group ranking to identify semantically meaningful feature groups, then introducing a learnable gating mechanism that jointly optimizes group-level relevance and perturbation masks. This approach is the first to integrate trainable group-level gating into counterfactual generation, significantly enhancing modality-group sparsity on the KneE-PAD dataset while matching or surpassing current baselines in terms of validity, temporal smoothness, and computational efficiency. The resulting explanations yield concise, actionable corrective guidance consistent with clinical practice.

Abstract

Counterfactual explanations (CEs) for multivariate time-series classifiers are often difficult to interpret in domains where experts reason in terms of semantic feature groups rather than individual channels. In rehabilitation movement analysis with multi-sensor inertial measurement units (IMUs), clinicians interpret motion through muscle-group and joint-segment abstractions; yet, most existing counterfactual methods operate at the channel level, producing scattered and biomechanically incoherent explanations. We propose a two-stage framework for group-based counterfactual generation in high-dimensional IMU data. We first show that Shapley-Adaptive (SA) group ranking preserves counterfactual validity but fails to enforce group-level sparsity, motivating the need for explicit group selection. We then introduce Learnable Gate (LG) methods, which incorporate trainable per-group relevance gates jointly optimized with perturbation masks. Experiments on the KneE-PAD rehabilitation dataset demonstrate that LG substantially improves modality-group sparsity compared to the channel-level M-CELS baseline while maintaining or improving validity, temporal smoothness, and generation efficiency. Exercise-specific analyses further show that group-structured counterfactuals yield concise, muscle-level corrective guidance aligned with clinical reasoning. Overall, the proposed framework enhances interpretability without sacrificing counterfactual quality, enabling more actionable explanations for rehabilitation movement analysis.

Problem

• counterfactual explanations
• time-series rehabilitation data
• feature groups
• interpretability
• inertial measurement units

Innovation

• group-based counterfactual
• learnable gate
• time-series explanation
• rehabilitation movement analysis
• modality-group sparsity