Multi-scale Frequency-Aware Adversarial Network for Parkinson's Disease Assessment Using Wearable Sensors

2025-10-12

AI Summary

To address the limitations of generic temporal models in Parkinson’s disease (PD) severity assessment—namely, insufficient pathology specificity and signal dilution caused by sparse symptomatic motion segments—this paper proposes a Multi-scale Frequency-domain Guided Adversarial Network (MF-GAN). The method introduces a medical-prior-driven frequency-domain decomposition module to enhance discriminability of PD-specific features, such as tremor and bradykinesia; integrates attention mechanisms with a multiple-instance learning framework to adaptively focus on diagnostically informative yet temporally sparse motion segments. By synergistically combining multi-scale frequency analysis, adversarial training, and temporal modeling, MF-GAN achieves significant improvements over state-of-the-art temporal models on both the public PADS dataset and a private four-class dataset: accuracy increases by 4.2% for binary PD classification and by 5.8% for multi-level severity assessment.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Multiagent Systems: Adversarial Agents
• Machine Learning: Multimodal Learning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Severity assessment of Parkinson's disease (PD) using wearable sensors offers an effective, objective basis for clinical management. However, general-purpose time series models often lack pathological specificity in feature extraction, making it difficult to capture subtle signals highly correlated with PD.Furthermore, the temporal sparsity of PD symptoms causes key diagnostic features to be easily "diluted" by traditional aggregation methods, further complicating assessment. To address these issues, we propose the Multi-scale Frequency-Aware Adversarial Multi-Instance Network (MFAM). This model enhances feature specificity through a frequency decomposition module guided by medical prior knowledge. Furthermore, by introducing an attention-based multi-instance learning (MIL) framework, the model can adaptively focus on the most diagnostically valuable sparse segments.We comprehensively validated MFAM on both the public PADS dataset for PD versus differential diagnosis (DD) binary classification and a private dataset for four-class severity assessment. Experimental results demonstrate that MFAM outperforms general-purpose time series models in handling complex clinical time series with specificity, providing a promising solution for automated assessment of PD severity.

Problem

• Improving pathological specificity in Parkinson's disease feature extraction
• Addressing temporal sparsity of symptoms in diagnostic assessment
• Enhancing automated severity classification using wearable sensor data

Innovation

• Multi-scale frequency decomposition guided by medical knowledge
• Attention-based multi-instance learning for sparse segments
• Adversarial network for Parkinson's disease severity assessment