Scholarlink

Deep learning-based dementia classification may have great potential for improving the early diagnosis and treatment outcomes of diseases. However, building a deep learning model with a high level of accuracy for medical applications is a challenging task due to limited clinical data. This results in the degradation of performance while identifying the existence of dementia. The limited availability of medical data often hinders the performance of any deep learning model, which restricts its use for real-world applications. The primary objective of this research is to enhance the classification of dementia by utilizing synthetic data augmentation with Deep Convolutional Generative Adversarial Networks (DCGANs). The synthesized images are used to increase the classification accuracy of Visual Geometry Group-19 (VGG-19). The dataset used for this research is brain MRI scans acquired from Kaggle. The GAN-generated images were added to the original dataset, and the combined dataset was used to train our models. The performance of the proposed DCGAN-VGG-19 is compared with DCGAN-VGG-16. Additionally, the performance of both models was evaluated without synthesized images. The accuracy of VGG-19 increased from 69.66 to 70.83% when using augmented images. From the results, it is observed that DCGAN-VGG-19 significantly improves the accuracy of detecting dementia.

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

Enhancing Classification of Dementia with Synthetic Data Augmentation Using Deep Convolutional Generative Adversarial Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sai Venkatesh Ponnuru,
  • Rajalakshmi Shenbaga Moorthy

摘要

Deep learning-based dementia classification may have great potential for improving the early diagnosis and treatment outcomes of diseases. However, building a deep learning model with a high level of accuracy for medical applications is a challenging task due to limited clinical data. This results in the degradation of performance while identifying the existence of dementia. The limited availability of medical data often hinders the performance of any deep learning model, which restricts its use for real-world applications. The primary objective of this research is to enhance the classification of dementia by utilizing synthetic data augmentation with Deep Convolutional Generative Adversarial Networks (DCGANs). The synthesized images are used to increase the classification accuracy of Visual Geometry Group-19 (VGG-19). The dataset used for this research is brain MRI scans acquired from Kaggle. The GAN-generated images were added to the original dataset, and the combined dataset was used to train our models. The performance of the proposed DCGAN-VGG-19 is compared with DCGAN-VGG-16. Additionally, the performance of both models was evaluated without synthesized images. The accuracy of VGG-19 increased from 69.66 to 70.83% when using augmented images. From the results, it is observed that DCGAN-VGG-19 significantly improves the accuracy of detecting dementia.