Scholarlink

<p>Globally, brain tumor is the leading disorder that affects an individual’s lifecycle. However, timely and accurate brain tumor diagnosis is important to increase the patient’s lifespan and improve personalized treatment decisions. Therefore, the existing studies implemented various deep-learning approaches to perform brain tumor classification. Nevertheless, none of the prevailing mechanisms concentrated on the red nucleus segmentation during brain tumor identification, diminishing the accuracy of brain tumor detection. Thus, this paper proposes a red nucleus region segmentation-based brain tumor classification and stage estimation using GLRU-SGNet and Deep-2C-MHONN. Primarily, the user registration is done, followed by key generation. Then, the IoMT data is encrypted using MHCC for security. At the receiver side, the IoMT data is decrypted using MHCC. Next, the IoMT data is subjected to the brain tumor diagnosis. Here, the brain MRI dataset is gathered and then pre-processed. Afterward, the with skull and without skull regions are split. From the without skull regions, B0 and B1 inhomogeneity artifact removal by R2NLM, image resolution enhancement, tissue clustering by K-means, core quotient analysis, and red nucleus segmentation by GLRU-SGNet are done. Likewise, from the with skull region, the epidural space is obtained. Based on the epidural space, segmented red nucleus, and resolution-enhanced image, the tumor region is segmented using W2CBS. Afterward, features are extracted from the segmented tumor region. Thereafter, by using Deep-2C-MHONN, the tumor is classified as normal, meningioma, glioma, and pituitary. Finally, the stages of the tumor are identified based on the Adaptive-Neuro Fuzzy Log-MaxAbs Scaling Inference System (ANF-LMASIS). Thus, the proposed work had higher superiority due to the inclusion of red nucleus region segmentation. The experimental outcomes proved that the proposed GLRU-SGNet obtained a low MAE (0.03) and high dice score (0.9872) for red nucleus segmentation, whereas the existing V-Net attained a high MAE (1.0303) and low dice score (0.2233). Based on the segmented red nucleus region, the proposed Deep-2C-MHONN classified the brain tumors with a high accuracy of 98.97%. Therefore, when compared to the baseline methods, the proposed model provided enhanced performance in a clinical setting.</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

Red nucleus region segmentation-based brain tumor classification and stage estimation using GLRN-SGNet and Deep-2C-MHONN

  • 期刊论文

  • 发表日期：

    2026年3月11日

查看全文
  • Vijay Kumar Gugulothu

摘要

Globally, brain tumor is the leading disorder that affects an individual’s lifecycle. However, timely and accurate brain tumor diagnosis is important to increase the patient’s lifespan and improve personalized treatment decisions. Therefore, the existing studies implemented various deep-learning approaches to perform brain tumor classification. Nevertheless, none of the prevailing mechanisms concentrated on the red nucleus segmentation during brain tumor identification, diminishing the accuracy of brain tumor detection. Thus, this paper proposes a red nucleus region segmentation-based brain tumor classification and stage estimation using GLRU-SGNet and Deep-2C-MHONN. Primarily, the user registration is done, followed by key generation. Then, the IoMT data is encrypted using MHCC for security. At the receiver side, the IoMT data is decrypted using MHCC. Next, the IoMT data is subjected to the brain tumor diagnosis. Here, the brain MRI dataset is gathered and then pre-processed. Afterward, the with skull and without skull regions are split. From the without skull regions, B0 and B1 inhomogeneity artifact removal by R2NLM, image resolution enhancement, tissue clustering by K-means, core quotient analysis, and red nucleus segmentation by GLRU-SGNet are done. Likewise, from the with skull region, the epidural space is obtained. Based on the epidural space, segmented red nucleus, and resolution-enhanced image, the tumor region is segmented using W2CBS. Afterward, features are extracted from the segmented tumor region. Thereafter, by using Deep-2C-MHONN, the tumor is classified as normal, meningioma, glioma, and pituitary. Finally, the stages of the tumor are identified based on the Adaptive-Neuro Fuzzy Log-MaxAbs Scaling Inference System (ANF-LMASIS). Thus, the proposed work had higher superiority due to the inclusion of red nucleus region segmentation. The experimental outcomes proved that the proposed GLRU-SGNet obtained a low MAE (0.03) and high dice score (0.9872) for red nucleus segmentation, whereas the existing V-Net attained a high MAE (1.0303) and low dice score (0.2233). Based on the segmented red nucleus region, the proposed Deep-2C-MHONN classified the brain tumors with a high accuracy of 98.97%. Therefore, when compared to the baseline methods, the proposed model provided enhanced performance in a clinical setting.