Scholarlink

<p>Accurate multi‑disease classification is vital for effective diagnosis and treatment. Conditions such as Alzheimer’s disease, which impairs memory and cognition, and brain tumors, often fatal and difficult to detect early, pose major challenges in clinical practice. Traditional models frequently delay precise diagnosis and treatment. To address these limitations, this study introduces the Fractional Convolutional Residual Network (FCR‑Net) for improved multi‑disease classification. Initially, the MRI image is forwarded as input to the pre-processing phase, where the noise is eradicated by utilizing a bilateral filter and a linear smoothing filter. Next, the affected region from an image is segmented by using ZNet, which is trained by Taylor Artificial Protozoa Optimization (TAPO). After this, the feature extraction is executed, and then the multi-disease classification is performed by exploiting FCR-Net. Additionally, this study utilized the Multi-Disease Dataset, which contains data spanning 28 categories of medical conditions, including keratosis, diabetic retinopathy, brain tumors, COVID-19, pneumonia, and several others. Furthermore, the FCR-Net approach computed a maximum True Negative Rate (TNR), Negative Predictive Value (NPV), accuracy, Positive Predictive Value (PPV), True Positive Rate (TPR) and F1-score of 93.535%, 92.698%, 95.478%, 92.440%, 94.997% and 93.090% respectively. Additionally, the proposed method improves the performance accuracy over the traditional methods, such as Multi-scale and Auto-tuned Semi-supervised Deep Subspace Clustering (MAS-DSC), NASNet-Mobile, Visual Geometry Group 16-Support Vector Machine (VGG16-SVM), Adaptive Hybrid Attention Network (AHANet), MobileNet-DenseNet and Transformer Attention Hybrid Network (MDTACNet), and Bilateral ConvNeXt-based Structure-aware Multi-scale Multi-view Learning Network (BiNeXt-SMSMVL) is 12.09%, 8.86%, 6.01%, 1.87%, 2.38%, and 1.18%.</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

Deep Learning-based Enhanced Fractional Convolutional Residual Network (FCR-Net) for Neurological Disorders Using MRI Images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月17日

查看全文
  • Anthuvan Lydia M,
  • Santhi M

摘要

Accurate multi‑disease classification is vital for effective diagnosis and treatment. Conditions such as Alzheimer’s disease, which impairs memory and cognition, and brain tumors, often fatal and difficult to detect early, pose major challenges in clinical practice. Traditional models frequently delay precise diagnosis and treatment. To address these limitations, this study introduces the Fractional Convolutional Residual Network (FCR‑Net) for improved multi‑disease classification. Initially, the MRI image is forwarded as input to the pre-processing phase, where the noise is eradicated by utilizing a bilateral filter and a linear smoothing filter. Next, the affected region from an image is segmented by using ZNet, which is trained by Taylor Artificial Protozoa Optimization (TAPO). After this, the feature extraction is executed, and then the multi-disease classification is performed by exploiting FCR-Net. Additionally, this study utilized the Multi-Disease Dataset, which contains data spanning 28 categories of medical conditions, including keratosis, diabetic retinopathy, brain tumors, COVID-19, pneumonia, and several others. Furthermore, the FCR-Net approach computed a maximum True Negative Rate (TNR), Negative Predictive Value (NPV), accuracy, Positive Predictive Value (PPV), True Positive Rate (TPR) and F1-score of 93.535%, 92.698%, 95.478%, 92.440%, 94.997% and 93.090% respectively. Additionally, the proposed method improves the performance accuracy over the traditional methods, such as Multi-scale and Auto-tuned Semi-supervised Deep Subspace Clustering (MAS-DSC), NASNet-Mobile, Visual Geometry Group 16-Support Vector Machine (VGG16-SVM), Adaptive Hybrid Attention Network (AHANet), MobileNet-DenseNet and Transformer Attention Hybrid Network (MDTACNet), and Bilateral ConvNeXt-based Structure-aware Multi-scale Multi-view Learning Network (BiNeXt-SMSMVL) is 12.09%, 8.86%, 6.01%, 1.87%, 2.38%, and 1.18%.