Scholarlink

<p>Early detection of Parkinson’s disease (PD) remains a major biomedical challenge, particularly in the early stages where symptoms are subtle and similar to those of healthy individuals. In this study, we propose a sequence-based classification framework using Bidirectional Gated Recurrent Units (BiGRUs) to exploit the temporal dynamics of handwriting signals for PD identification. The proposed architecture processes both raw and derived feature sequences, followed by dense layers for final classification. Experiments conducted on the NewHandPD dataset demonstrate strong performance, achieving an overall accuracy above 94% and a precision of approximately 95% across multiple handwriting tasks. Furthermore, statistical validation using the Friedman and Nemenyi tests confirms that the proposed method significantly outperforms benchmark approaches. These findings highlight the effectiveness of sequence-based deep learning models for early PD detection through dynamic handwriting analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic Handwriting Analysis Using Sequence-Based on BiGRUs for Parkinson’s Disease Identification

  • 期刊论文

  • 发表日期：

    2026年3月21日

查看全文
  • Houssam Moustansir,
  • Aissam Hadri,
  • Lekbir Afraites,
  • Soufiane Lyaqini

摘要

Early detection of Parkinson’s disease (PD) remains a major biomedical challenge, particularly in the early stages where symptoms are subtle and similar to those of healthy individuals. In this study, we propose a sequence-based classification framework using Bidirectional Gated Recurrent Units (BiGRUs) to exploit the temporal dynamics of handwriting signals for PD identification. The proposed architecture processes both raw and derived feature sequences, followed by dense layers for final classification. Experiments conducted on the NewHandPD dataset demonstrate strong performance, achieving an overall accuracy above 94% and a precision of approximately 95% across multiple handwriting tasks. Furthermore, statistical validation using the Friedman and Nemenyi tests confirms that the proposed method significantly outperforms benchmark approaches. These findings highlight the effectiveness of sequence-based deep learning models for early PD detection through dynamic handwriting analysis.