Scholarlink

We propose a dual branch Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) model to differentiate between Parkinson’s disease (PD) patients and healthy subjects using gait data. Spatial and temporal features are extracted from the left and right feet gait data. The obtained feature map is combined using an LSTM layer, and the attention layer is used to concentrate features specific to a time zone. We use a publicly available PhysioNet data, which contains the vertical ground reaction force of the left and right feet of subjects. Results demonstrate that the proposed model performs better than existing techniques, giving a test set accuracy of 99.3%. Additional experiments are conducted to find out the best suitable time step, which again helps to increase the test set accuracy to 99.7%.

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

A Dual Branch Attention Enhanced CNN-LSTM Model for Diagnosis of Parkinson’s Disease

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rakesh Kiran,
  • Deepak Sharma,
  • Ashish Anand

摘要

We propose a dual branch Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) model to differentiate between Parkinson’s disease (PD) patients and healthy subjects using gait data. Spatial and temporal features are extracted from the left and right feet gait data. The obtained feature map is combined using an LSTM layer, and the attention layer is used to concentrate features specific to a time zone. We use a publicly available PhysioNet data, which contains the vertical ground reaction force of the left and right feet of subjects. Results demonstrate that the proposed model performs better than existing techniques, giving a test set accuracy of 99.3%. Additional experiments are conducted to find out the best suitable time step, which again helps to increase the test set accuracy to 99.7%.