MOSAIC: Modality-Specific Adaptation for Incremental Continual Learning in Parkinson's Disease Gait Assessment

2026-06-11

AI Summary

This work addresses the challenges of modality-incremental learning in Parkinson’s disease gait assessment, where sensor heterogeneity, data privacy constraints, and storage limitations lead to unreliable cross-modal distillation, modality-specific statistical shifts, and degraded model plasticity. To overcome these issues, the authors propose the MOSAIC framework, which stabilizes representations of newly introduced modalities through modality-specific warm-up, employs a statistically decoupled multimodal batch normalization (MSBN) architecture, and incorporates a curriculum-guided repulsion objective to restore model plasticity. Notably, MOSAIC is the first to identify and mitigate the “toxic teacher” problem in this context. Evaluated on three multimodal Parkinson’s gait datasets, the method achieves superior final performance, significantly alleviates catastrophic forgetting, and outperforms existing continual learning approaches.

Abstract

Gait-based Parkinson's disease assessment increasingly relies on heterogeneous sensors, but clinical systems rarely collect all modalities simultaneously. New sensors may arrive through device upgrades, protocol changes, or multi-center deployment, while historical patient data are often unavailable because of privacy and storage constraints. This modality-incremental setting faces three challenges: unreliable cross-modal distillation, modality-specific statistical shifts, and reduced plasticity after preservation. We propose MOSAIC, a compact continual learning framework. First, we identify the Toxic Teacher phenomenon and introduce Modality-Specific Warm-Up to stabilize newly learned modality representations before distillation. Second, we propose a statistics-decoupled MSBN architecture that isolates sensor statistics while maintaining a shared semantic backbone. Third, we design a curriculum-guided repulsive objective for Plasticity Recovery, preserving legacy knowledge while recovering modality-specific capacity. Experiments on three multimodal Parkinson's gait datasets show that MOSAIC improves final performance and mitigates forgetting. Project code is available at: https://github.com/minlinzeng/MOSAIC_Modality-Specific-Adaptation-for-Incremental-Continual-Learning-in-PD-Gait-Assessment.git

Problem

• modality-incremental learning
• Parkinson's disease gait assessment
• continual learning
• cross-modal distillation
• statistical shift

Innovation

• Modality-Incremental Learning
• Continual Learning
• Modality-Specific Adaptation
• Statistical Decoupling
• Plasticity Recovery