Scholarlink

<p>Identifying Rheumatoid Arthritis (RA) is crucial for preventing irreversible joint damage as well as functional impairment by facilitating early detection and swift intervention. Progress in medical imaging as well as Artificial Intelligence (AI)&#xa0;has facilitated automated RA recognition from X-ray, MRI, and ultrasound images, enhancing diagnostic reliability and effectiveness. Nonetheless, obstacles persist because of subtle early-stage symptoms, variations in joint appearance, imaging noise, constrained annotated datasets, as well as class imbalance, all of which impede reliable and generalized detection effectiveness. Hence, this paper performs the Rheumatoid Arthritis Detection (RAD) model using intelligent deep learning methodology. The dataset is initially collected from standard benchmark sources such as Synapse named “Rheumatoid Arthritis 2-Dialogue for Reverse Engineering Assessment and Methods (RA2-DREAM)”. The pre-processing is next done using the Edge-Preserving Diffusion Filtering (EPDF) technique. Further, the segmentation is followed with the help of UNet++ approach. The extraction of the features is done by the Medical Image Attention-based Feature Extractor (MIAFEx) technique. Finally, the detection of the proposed RAD model is accomplished using the novel Enhanced Recurrent Neural Network (ERNN) model. The parameter optimization in RNN is performed by the nature inspired optimization algorithm referred as Hiking Optimization Algorithm (HOA), which returns the accuracy maximization as the fitness function. On the independent test set, the proposed ERNN-HOA achieved an accuracy of 98.20% (95% CI 97.12–99.02), specificity of 96.80%, precision of 97.50%, sensitivity of 97.65%, F1-score of 97.85%, and an AUROC of 0.992 (95% CI 0.987–0.996). These results demonstrate statistically robust and clinically reliable performance, significantly outperforming existing comparative models for automated rheumatoid arthritis detection.</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

Rheumatoid arthritis detection using enhanced recurrent neural network-based optimization model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月9日

查看全文
  • S. Vilma Veronica,
  • G. Bhuvaneswari

摘要

Identifying Rheumatoid Arthritis (RA) is crucial for preventing irreversible joint damage as well as functional impairment by facilitating early detection and swift intervention. Progress in medical imaging as well as Artificial Intelligence (AI) has facilitated automated RA recognition from X-ray, MRI, and ultrasound images, enhancing diagnostic reliability and effectiveness. Nonetheless, obstacles persist because of subtle early-stage symptoms, variations in joint appearance, imaging noise, constrained annotated datasets, as well as class imbalance, all of which impede reliable and generalized detection effectiveness. Hence, this paper performs the Rheumatoid Arthritis Detection (RAD) model using intelligent deep learning methodology. The dataset is initially collected from standard benchmark sources such as Synapse named “Rheumatoid Arthritis 2-Dialogue for Reverse Engineering Assessment and Methods (RA2-DREAM)”. The pre-processing is next done using the Edge-Preserving Diffusion Filtering (EPDF) technique. Further, the segmentation is followed with the help of UNet++ approach. The extraction of the features is done by the Medical Image Attention-based Feature Extractor (MIAFEx) technique. Finally, the detection of the proposed RAD model is accomplished using the novel Enhanced Recurrent Neural Network (ERNN) model. The parameter optimization in RNN is performed by the nature inspired optimization algorithm referred as Hiking Optimization Algorithm (HOA), which returns the accuracy maximization as the fitness function. On the independent test set, the proposed ERNN-HOA achieved an accuracy of 98.20% (95% CI 97.12–99.02), specificity of 96.80%, precision of 97.50%, sensitivity of 97.65%, F1-score of 97.85%, and an AUROC of 0.992 (95% CI 0.987–0.996). These results demonstrate statistically robust and clinically reliable performance, significantly outperforming existing comparative models for automated rheumatoid arthritis detection.