Scholarlink

<p>Early and accurate diagnosis of Parkinson’s disease (PD) is crucial but challenging due to the limited availability of labelled MRI data and the disease’s inherent variability. The use of traditional deep learning models faces critical limitations which create obstacles for their acceptance in clinical practice including overfitting problems and difficulty in interpretation. To overcome these limitations in this, article introduced ACT Net which represents an attentive CNN-transformer hybrid model that utilizes the best characteristics of convolutional neural networks (CNNs) and transformers. ACT Net advances the performance of MRI image-based models through its capability to identify local and global features from the images and through its generative AI augmentation of limited data supplies which enhances model robustness. The implementation of explainable AI (XAI) methods delivers understandable explanations about model decisions to healthcare professionals for building their trust. ACT Net reaches a 96.2% accuracy level for classification that establishes better results than conventional methods both in terms of diagnostic precision and interpretability of the predictive model. The combination of addressing important challenges regarding scarce data and modelling clarity positions ACT Net as a powerful system for early PD diagnosis and improved patient care.</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

Parkinson’s disease early detection using hybrid attentive CNN-transformer model

  • 期刊论文

  • 发表日期：

    2025年9月14日

查看全文
  • G. Rohini Phaneendra Kumari,
  • M. Ravi Kanth,
  • M. V. Kamal

摘要

Early and accurate diagnosis of Parkinson’s disease (PD) is crucial but challenging due to the limited availability of labelled MRI data and the disease’s inherent variability. The use of traditional deep learning models faces critical limitations which create obstacles for their acceptance in clinical practice including overfitting problems and difficulty in interpretation. To overcome these limitations in this, article introduced ACT Net which represents an attentive CNN-transformer hybrid model that utilizes the best characteristics of convolutional neural networks (CNNs) and transformers. ACT Net advances the performance of MRI image-based models through its capability to identify local and global features from the images and through its generative AI augmentation of limited data supplies which enhances model robustness. The implementation of explainable AI (XAI) methods delivers understandable explanations about model decisions to healthcare professionals for building their trust. ACT Net reaches a 96.2% accuracy level for classification that establishes better results than conventional methods both in terms of diagnostic precision and interpretability of the predictive model. The combination of addressing important challenges regarding scarce data and modelling clarity positions ACT Net as a powerful system for early PD diagnosis and improved patient care.