Scholarlink

Purpose <p>Dementia, particularly Alzheimer’s disease (AD), is a leading cause of disability and dependency among the elderly. Alzheimer’s accounts for approximately 75% of dementia cases in Taiwan and is marked by early hippocampal atrophy and memory decline. This study aimed to develop a deep learning-based system for early AD prediction using brain MRI data for enhanced diagnostic support in clinical settings.</p> Methods <p>A two-stage model is proposed based on the 214 brain MR images from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). First, a U-Net architecture was employed to perform automated segmentation of the hippocampus. Subsequently, the segmented images were classified using a ResNet-101 model to predict the presence of AD.</p> Results <p>The hippocampus segmentation model achieved a Dice similarity coefficient of 0.886 and an intersection over union (IoU) of 0.795, indicating precise region delineation. The AD prediction model yielded an accuracy of 0.833, precision of 0.847, and recall of 0.863, demonstrating robust performance in early-stage AD classification.</p> Conclusion <p>The proposed system integrates hippocampal segmentation and deep learning-based classification for accurate AD prediction. Its strong performance highlights its potential as a clinical tool for early detection and intervention in Alzheimer’s disease, contributing to improved patient outcomes and disease management.</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

Enhanced Hippocampus Segmentation and Alzheimer’s Disease Prediction Using U-Net and ResNet Models on Brain Magnetic Resonance Imaging

  • 期刊论文

  • 发表日期：

    2025年8月28日

查看全文
  • Ting-An Chang,
  • Chung-Wen Hung,
  • Xue-Yan Liu

摘要

Purpose

Dementia, particularly Alzheimer’s disease (AD), is a leading cause of disability and dependency among the elderly. Alzheimer’s accounts for approximately 75% of dementia cases in Taiwan and is marked by early hippocampal atrophy and memory decline. This study aimed to develop a deep learning-based system for early AD prediction using brain MRI data for enhanced diagnostic support in clinical settings.

Methods

A two-stage model is proposed based on the 214 brain MR images from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). First, a U-Net architecture was employed to perform automated segmentation of the hippocampus. Subsequently, the segmented images were classified using a ResNet-101 model to predict the presence of AD.

Results

The hippocampus segmentation model achieved a Dice similarity coefficient of 0.886 and an intersection over union (IoU) of 0.795, indicating precise region delineation. The AD prediction model yielded an accuracy of 0.833, precision of 0.847, and recall of 0.863, demonstrating robust performance in early-stage AD classification.

Conclusion

The proposed system integrates hippocampal segmentation and deep learning-based classification for accurate AD prediction. Its strong performance highlights its potential as a clinical tool for early detection and intervention in Alzheimer’s disease, contributing to improved patient outcomes and disease management.