Scholarlink

<p>Recently, Alzheimer’s disease (AD) has become a serious hazard to human health.&#xa0;Therefore, an optimal strategy for formulating the treatment plan is the AD’s early diagnosis. In spite of this, no effective treatment or accurate diagnosis exists currently. Also, the pre-selection of brain regions is a complicated task. Therefore, an efficient AD classification is needed. Hence, by utilizing the Exponential Gaussian Error Linear Unit–Squeeze Net (EGELU-SZN) technique, an early prediction as well as classification of AD is proposed. Primarily, from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset, the input brain Magnetic Resonance Imaging (MRI) images are gathered. For the brain MRI images, a skull stripping process is executed. Here, by utilizing the Adaptive Median Otsu’s Thresholding (AMOT) technique, the skull along with cerebral tissues like fat and skin around the brain are removed. After that, to eliminate the noise, pre-processing is computed. After pre-processing, to segment the brain region accurately, segmentation is evaluated. Therefore, an effectual segmentation algorithm termed Rectilinear Mayfly Optimization-centric Automatic Seeded Region Growing algorithm (RMF-ASRG) has been utilized. Features are extracted as of the segmented brain region. Later, by utilizing the Reflective Correlation Principal Component Analysis (RCPCA) algorithm, the reduction of features is performed. Then, the predicted outcomes are classified as AD, Cognitive Normal (CN), as well as Mild Cognitive Impairment (MCI) by employing the EGELU-SZN Classifier. The proposed EGELU-SZN attained the accuracy, precision, recall, specificity, and sensitivity values of 95.9882%, 94.1661%, 93.2327%, 92.8744%, and 96.2327%, respectively, in the classification process. Experimental outcomes signified that superior performance was attained by the proposed methodology when analyzed with benchmark methodologies.</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

An effective deep learning-based automatic prediction and classification of Alzheimer's disease using EGELU-SZN technique

  • 期刊论文

  • 发表日期：

    2025年1月27日

查看全文
  • B. Sathyabhama,
  • M. Kannan

摘要

Recently, Alzheimer’s disease (AD) has become a serious hazard to human health. Therefore, an optimal strategy for formulating the treatment plan is the AD’s early diagnosis. In spite of this, no effective treatment or accurate diagnosis exists currently. Also, the pre-selection of brain regions is a complicated task. Therefore, an efficient AD classification is needed. Hence, by utilizing the Exponential Gaussian Error Linear Unit–Squeeze Net (EGELU-SZN) technique, an early prediction as well as classification of AD is proposed. Primarily, from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset, the input brain Magnetic Resonance Imaging (MRI) images are gathered. For the brain MRI images, a skull stripping process is executed. Here, by utilizing the Adaptive Median Otsu’s Thresholding (AMOT) technique, the skull along with cerebral tissues like fat and skin around the brain are removed. After that, to eliminate the noise, pre-processing is computed. After pre-processing, to segment the brain region accurately, segmentation is evaluated. Therefore, an effectual segmentation algorithm termed Rectilinear Mayfly Optimization-centric Automatic Seeded Region Growing algorithm (RMF-ASRG) has been utilized. Features are extracted as of the segmented brain region. Later, by utilizing the Reflective Correlation Principal Component Analysis (RCPCA) algorithm, the reduction of features is performed. Then, the predicted outcomes are classified as AD, Cognitive Normal (CN), as well as Mild Cognitive Impairment (MCI) by employing the EGELU-SZN Classifier. The proposed EGELU-SZN attained the accuracy, precision, recall, specificity, and sensitivity values of 95.9882%, 94.1661%, 93.2327%, 92.8744%, and 96.2327%, respectively, in the classification process. Experimental outcomes signified that superior performance was attained by the proposed methodology when analyzed with benchmark methodologies.