Scholarlink

Alzheimer’s disease is affecting millions of people worldwide and causing loss of cognitive functions primarily in older individuals. Magnetic Resonance Imaging (MRI) facilitates disease screening by revealing changes in different brain regions. Early diagnosis is critical for treatment success and improving patients’ quality of life. MRI analysis is widely used for early Alzheimer’s diagnosis, enabling classification of the disease stages. Current diagnostic methods often rely on limited datasets, necessitating bigger datasets to enhance diagnostic performance. In this study, the unsupervised transition from image to image between two classes of MR images, namely those with no Alzheimer’s symptoms and those with Alzheimer’s symptoms, was investigated. Bidirectional synthetic data was generated by using Cycle Generative Adversarial Networks (CycleGAN). Synthetic data augmentation was performed by converting MRI images without Alzheimer’s symptoms into images with Alzheimer’s symptoms and performing the inverse transformation, and 100 MRI images of each class were generated. The performances of transfer learning-based binary classification on the original dataset and the dataset extended with CycleGAN were demonstrated. Performance evaluation has been performed with and without data augmentation. Performance improvements were observed for the dataset extended with CycleGAN compared to the original dataset.

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

Computer-Aided Alzheimer’s Disease Diagnosis from Magnetic Resonance Images Using Cycle Generative Adversarial Networks and Deep Transfer Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Seda Coşkun Eliküçük,
  • Volkan Müjdat Tiryaki

摘要

Alzheimer’s disease is affecting millions of people worldwide and causing loss of cognitive functions primarily in older individuals. Magnetic Resonance Imaging (MRI) facilitates disease screening by revealing changes in different brain regions. Early diagnosis is critical for treatment success and improving patients’ quality of life. MRI analysis is widely used for early Alzheimer’s diagnosis, enabling classification of the disease stages. Current diagnostic methods often rely on limited datasets, necessitating bigger datasets to enhance diagnostic performance. In this study, the unsupervised transition from image to image between two classes of MR images, namely those with no Alzheimer’s symptoms and those with Alzheimer’s symptoms, was investigated. Bidirectional synthetic data was generated by using Cycle Generative Adversarial Networks (CycleGAN). Synthetic data augmentation was performed by converting MRI images without Alzheimer’s symptoms into images with Alzheimer’s symptoms and performing the inverse transformation, and 100 MRI images of each class were generated. The performances of transfer learning-based binary classification on the original dataset and the dataset extended with CycleGAN were demonstrated. Performance evaluation has been performed with and without data augmentation. Performance improvements were observed for the dataset extended with CycleGAN compared to the original dataset.