Scholarlink

Aim <p>Currently, PD is diagnosed based on clinical symptoms, which can be difficult to identify the disease in its early stages. The main objective of this work is to enhance the accuracy and sensitivity of PD diagnosis using Smoothed Pseudo Wigner-Ville Distribution (SPWVD), combined with pre-trained convolutional neural network (CNN) models.</p> Methods <p>The key innovation of this work includes Time–Frequency Representation (TFR) techniques like Scalogram, Spectrogram, and Smoothed Pseudo Wigner-Ville Distribution (SPWVD), integrated with Inceptionv3, VGG16, and Mobilenetv2 models to classify PD patients from healthy control subjects effectively. The methodology involves collecting gait data from 93 Parkinson’s patients and 73 healthy control subjects and processing them to generate Time Frequency Representation (TFR) images. These images are given as inputs to pre-trained CNN models such as MobileNetv2, Inceptionv3, and VGG16. The proposed models are trained and tested using the gait data collected from physionet database.</p> Results <p>Finally, the MobileNetv2 pre-trained model on these SPWVD yields a highest accuracy of 99% during classification process. Performance metrics such as accuracy, Sensitivity, Specificity, Matthew Correlation Coefficient, and F1-score are used to evaluate the model.</p> Conclusions <p>The proposed method employing SPWVD and CNN models outperforms traditional diagnostic approaches by providing a faster and more precise approach. This could significantly improve early diagnosis and treatment initiation, potentially altering the progression of PD.</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

An efficient Parkinson's disease detection using smoothed pseudo-Wigner Ville distribution and MobileNetV2 convolutional neural network

  • 期刊论文

  • 发表日期：

    2025年4月14日

查看全文
  • Amaladass P. Klinton,
  • Priya S. Jeba,
  • George S. Thomas,
  • M. S. P. Subathra,
  • Ebenezer Shamila,
  • A. Ananthi,
  • Robertas Damaševičius

摘要

Aim

Currently, PD is diagnosed based on clinical symptoms, which can be difficult to identify the disease in its early stages. The main objective of this work is to enhance the accuracy and sensitivity of PD diagnosis using Smoothed Pseudo Wigner-Ville Distribution (SPWVD), combined with pre-trained convolutional neural network (CNN) models.

Methods

The key innovation of this work includes Time–Frequency Representation (TFR) techniques like Scalogram, Spectrogram, and Smoothed Pseudo Wigner-Ville Distribution (SPWVD), integrated with Inceptionv3, VGG16, and Mobilenetv2 models to classify PD patients from healthy control subjects effectively. The methodology involves collecting gait data from 93 Parkinson’s patients and 73 healthy control subjects and processing them to generate Time Frequency Representation (TFR) images. These images are given as inputs to pre-trained CNN models such as MobileNetv2, Inceptionv3, and VGG16. The proposed models are trained and tested using the gait data collected from physionet database.

Results

Finally, the MobileNetv2 pre-trained model on these SPWVD yields a highest accuracy of 99% during classification process. Performance metrics such as accuracy, Sensitivity, Specificity, Matthew Correlation Coefficient, and F1-score are used to evaluate the model.

Conclusions

The proposed method employing SPWVD and CNN models outperforms traditional diagnostic approaches by providing a faster and more precise approach. This could significantly improve early diagnosis and treatment initiation, potentially altering the progression of PD.