Scholarlink

<p>Brain tumor (BT) is a rapid growth of brain cells. If the BT is not identified and treated in the first stage, it could cause death. Despite several methods and efforts being developed for segmenting and identifying BT, the detection of BT is complicated due to the distinct position of the tumor and its size. To solve such issues, this paper proposes the Conditional Autoregressive Value-at-Risk_Xception Maxout-Network (Caviar_XM-Net) for BT detection utilizing magnetic resonance imaging (MRI) images. The input MRI image gathered from the dataset is denoised using the adaptive bilateral filter (ABF), and tumor region segmentation is done using BFC-MRFNet-RVSeg. Here, the segmentation is done by the Bayesian fuzzy clustering (BFC) and multi-branch residual fusion network (MRF-Net) separately. Subsequently, outputs from both segmentation techniques are combined using the RV coefficient. Image augmentation is performed to boost the quantity of images in the training process. Afterwards, feature extraction is done, where features, like local optimal oriented pattern (LOOP), convolutional neural network (CNN) features, median binary pattern (MBP) with statistical features, and local Gabor XOR pattern (LGXP), are extracted. Lastly, BT detection is carried out by employing Caviar_XM-Net, which is acquired by the assimilation of the Xception model and deep Maxout network (DMN) with the CAViaR approach. Furthermore, the effectiveness of Caviar_XM-Net is examined using the parameters, namely sensitivity, accuracy, specificity, precision, and F1-score, and the corresponding values of 91.59%, 91.36%, 90.83%, 90.99%, and 91.29% are attained. Hence, the Caviar_XM-Net performs better than the traditional methods with high efficiency.</p> Graphical abstract <p>In this article, Caviar_XM-Net is introduced for BT detection utilizing MRI images. Originally, input MRI image is attained from given database. After that, image denoising is performed using the ABF, where tumor region segmentation was done using BFC and MRF-Net separately. Subsequently, the outputs from both segmentation techniques are combined using the RV coefficient. Then, image augmentation is carried out where translation, contrast, color augmentation, scaling, and rotation are exploited to enhance the amount of images for training procedure. Thereafter, feature extraction is executed, where features, like CNN features, MBP with statistical features, LOOP, and LGXP, are extracted. Finally, BT detection is accomplished using the Caviar_XM-Net, which is obtained by the combination of the Xception model and DMN with the CAViaR approach, where layers are modified.</p> <p></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

Hybrid segmentation model and CAViaR -based Xception Maxout network for brain tumor detection using MRI images

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • Satha Swapna,
  • Yugandhar Garapati

摘要

Brain tumor (BT) is a rapid growth of brain cells. If the BT is not identified and treated in the first stage, it could cause death. Despite several methods and efforts being developed for segmenting and identifying BT, the detection of BT is complicated due to the distinct position of the tumor and its size. To solve such issues, this paper proposes the Conditional Autoregressive Value-at-Risk_Xception Maxout-Network (Caviar_XM-Net) for BT detection utilizing magnetic resonance imaging (MRI) images. The input MRI image gathered from the dataset is denoised using the adaptive bilateral filter (ABF), and tumor region segmentation is done using BFC-MRFNet-RVSeg. Here, the segmentation is done by the Bayesian fuzzy clustering (BFC) and multi-branch residual fusion network (MRF-Net) separately. Subsequently, outputs from both segmentation techniques are combined using the RV coefficient. Image augmentation is performed to boost the quantity of images in the training process. Afterwards, feature extraction is done, where features, like local optimal oriented pattern (LOOP), convolutional neural network (CNN) features, median binary pattern (MBP) with statistical features, and local Gabor XOR pattern (LGXP), are extracted. Lastly, BT detection is carried out by employing Caviar_XM-Net, which is acquired by the assimilation of the Xception model and deep Maxout network (DMN) with the CAViaR approach. Furthermore, the effectiveness of Caviar_XM-Net is examined using the parameters, namely sensitivity, accuracy, specificity, precision, and F1-score, and the corresponding values of 91.59%, 91.36%, 90.83%, 90.99%, and 91.29% are attained. Hence, the Caviar_XM-Net performs better than the traditional methods with high efficiency.

Graphical abstract

In this article, Caviar_XM-Net is introduced for BT detection utilizing MRI images. Originally, input MRI image is attained from given database. After that, image denoising is performed using the ABF, where tumor region segmentation was done using BFC and MRF-Net separately. Subsequently, the outputs from both segmentation techniques are combined using the RV coefficient. Then, image augmentation is carried out where translation, contrast, color augmentation, scaling, and rotation are exploited to enhance the amount of images for training procedure. Thereafter, feature extraction is executed, where features, like CNN features, MBP with statistical features, LOOP, and LGXP, are extracted. Finally, BT detection is accomplished using the Caviar_XM-Net, which is obtained by the combination of the Xception model and DMN with the CAViaR approach, where layers are modified.