Scholarlink

Alzheimer’s disease (AD) is a degenerative brain condition that impairs cognitive function, erodes memory, and alters behavior. The key to managing this disease lies in early detection. This study introduces a cutting-edge deep learning framework for detecting AD using Magnetic resonance imaging (MRI) data. In this paper, we have taken MRI images of both dementia and non-dementia subjects. We presented two architectures: a Convolutional Neural Network (CNN) and a Hybrid model. The hybrid model combines CNN and LSTM layers to analyze sequential dependencies in the data. We have done a comparison analysis between these two models. Our results show that the hybrid CNN-LSTM model outperforms the CNN model, achieving a classification accuracy of 98.70% compared to 98% for the CNN model. Besides, we have done a comparative study of ROC-AUC curve and confusion matrix to show the execution of the model. Again, at training procedure, model loss and accuracy also depicted according to epoch times. The researches already performed on prediction of Alzheimer’s disease has given poor performance of 94% classification accuracy in comparison to our model.

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

Deep Network Models for Alzheimer’s Diagnosis Using MRI: A Comparative Analysis of CNN and Hybrid CNN-LSTM Architectures

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sanghamitra Panda,
  • Vikas Singh,
  • Sushree Sasmita Dash,
  • Ashis Kumar Ratha,
  • Sushanta Meher,
  • Debasish Swapnesh Kumar Nayak

摘要

Alzheimer’s disease (AD) is a degenerative brain condition that impairs cognitive function, erodes memory, and alters behavior. The key to managing this disease lies in early detection. This study introduces a cutting-edge deep learning framework for detecting AD using Magnetic resonance imaging (MRI) data. In this paper, we have taken MRI images of both dementia and non-dementia subjects. We presented two architectures: a Convolutional Neural Network (CNN) and a Hybrid model. The hybrid model combines CNN and LSTM layers to analyze sequential dependencies in the data. We have done a comparison analysis between these two models. Our results show that the hybrid CNN-LSTM model outperforms the CNN model, achieving a classification accuracy of 98.70% compared to 98% for the CNN model. Besides, we have done a comparative study of ROC-AUC curve and confusion matrix to show the execution of the model. Again, at training procedure, model loss and accuracy also depicted according to epoch times. The researches already performed on prediction of Alzheimer’s disease has given poor performance of 94% classification accuracy in comparison to our model.