Scholarlink

Purpose <p>Medical imaging such as magnetic resonance imaging (MRI) and positron emission tomography (PET) has been combined with deep learning (DL) for improving the classification of healthy control (HC), mild cognitive impairment (MCI), and Alzheimer’s disease (AD). Compared to MRI and PET, computed tomography (CT) is a relatively cheap and fast imaging technique, which, however, is rarely used for the classification of HC/MCI/AD.</p> Methods <p>In this study, we investigate the feasibility of combining CT images with DL for the classification of HC/MCI/AD. A total of 1088 subjects including 483 AD patients, 314 MCI patients, and 291 HC subjects were enrolled in the study. Three-dimensional (3D) CT images were pre-processed by skull stripping, spatial normalization, and a brain intensity window. We trained several DL models using 3D CT images, and used an ensemble learning model to improve the classification of HC/MCI/AD.</p> Results <p>Our results showed that the three-class classification accuracy of the ensemble learning model utilizing 3D CT images was only 58.9%. However, the ensemble learning model achieved an accuracy of 83.3%, sensitivity of 88.3% and specificity of 73.3% in discriminating MCI/AD from HC.</p> Conclusion <p>The experimental results suggest that CT images combining with DL cannot be used as a diagnostic tool for the three-way classification task (HC vs. MCI vs. AD), but has the potential to be a screening tool for discriminating MCI/AD from HC.</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

Classification of HC, MCI, and AD Based on CT Using Deep Learning

  • 期刊论文

  • 发表日期：

    2025年10月16日

查看全文
  • Ing-Tsung Hsiao,
  • Kun-Ju Lin,
  • Chiung-Chih Chang,
  • Hsuan-Ming Huang

摘要

Purpose

Medical imaging such as magnetic resonance imaging (MRI) and positron emission tomography (PET) has been combined with deep learning (DL) for improving the classification of healthy control (HC), mild cognitive impairment (MCI), and Alzheimer’s disease (AD). Compared to MRI and PET, computed tomography (CT) is a relatively cheap and fast imaging technique, which, however, is rarely used for the classification of HC/MCI/AD.

Methods

In this study, we investigate the feasibility of combining CT images with DL for the classification of HC/MCI/AD. A total of 1088 subjects including 483 AD patients, 314 MCI patients, and 291 HC subjects were enrolled in the study. Three-dimensional (3D) CT images were pre-processed by skull stripping, spatial normalization, and a brain intensity window. We trained several DL models using 3D CT images, and used an ensemble learning model to improve the classification of HC/MCI/AD.

Results

Our results showed that the three-class classification accuracy of the ensemble learning model utilizing 3D CT images was only 58.9%. However, the ensemble learning model achieved an accuracy of 83.3%, sensitivity of 88.3% and specificity of 73.3% in discriminating MCI/AD from HC.

Conclusion

The experimental results suggest that CT images combining with DL cannot be used as a diagnostic tool for the three-way classification task (HC vs. MCI vs. AD), but has the potential to be a screening tool for discriminating MCI/AD from HC.