Scholarlink

<p>Brain tumours (BT) are potentially fatal conditions that can severely impair patient quality of life and interfere with the function of the brain. Detection and identification at an early stage are essential for successful treatment. Deep learning techniques for image analysis and identification have become more popular among medical practitioners, producing more reliable and accurate outcomes. In order to identify brain malignancies, segmentation is the process of differentiating between healthy and aberrant brain tissues or cells. To address these issues, a novel Attention V-Net with RegNet (AVR-Net) has been proposed for efficient BT detection utilizing a multi-modality MRI image. Initially, an adaptive median filter (AMF) is utilized to preprocess the multimodality images (T1, T2, FLAIR) to remove noise artifacts. The DL-based Dense neural network (DNN) is used for extracting the relevant features and classifying the multimodal images as normal and abnormal. The tumorous images are fed into the Attention V-Net for the segmentation of tumours. Afterwards, ROI-based RegNet is used for detecting the normal-benign, benign, benign-malignant, and malignant from the multi-modal MRI images. The evaluation of the proposed AVR-Net achieves a total accuracy (ACC) is 99.24% based on the gathered dataset. The proposed attention V-Net attains the ACC was increased by 10.90%, 8.62%, 6.11%, and 3.58% over SegNet, U-Net, Attention U-Net and V-Net, respectively. The proposed AVR-Net enhances the overall ACC by 0.28%, 2.25%, and 13.34% for ResNet50, MobileNetV1, and KNN respectively.</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

Brain Tumour Detection and Segmentation via Region-of-Interest Aided Deep Learning Model

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • K. P. Ajitha Gladis,
  • M. Anlin Sahaya Infant Tinu,
  • R. Suguna,
  • V. PadmaJothi

摘要

Brain tumours (BT) are potentially fatal conditions that can severely impair patient quality of life and interfere with the function of the brain. Detection and identification at an early stage are essential for successful treatment. Deep learning techniques for image analysis and identification have become more popular among medical practitioners, producing more reliable and accurate outcomes. In order to identify brain malignancies, segmentation is the process of differentiating between healthy and aberrant brain tissues or cells. To address these issues, a novel Attention V-Net with RegNet (AVR-Net) has been proposed for efficient BT detection utilizing a multi-modality MRI image. Initially, an adaptive median filter (AMF) is utilized to preprocess the multimodality images (T1, T2, FLAIR) to remove noise artifacts. The DL-based Dense neural network (DNN) is used for extracting the relevant features and classifying the multimodal images as normal and abnormal. The tumorous images are fed into the Attention V-Net for the segmentation of tumours. Afterwards, ROI-based RegNet is used for detecting the normal-benign, benign, benign-malignant, and malignant from the multi-modal MRI images. The evaluation of the proposed AVR-Net achieves a total accuracy (ACC) is 99.24% based on the gathered dataset. The proposed attention V-Net attains the ACC was increased by 10.90%, 8.62%, 6.11%, and 3.58% over SegNet, U-Net, Attention U-Net and V-Net, respectively. The proposed AVR-Net enhances the overall ACC by 0.28%, 2.25%, and 13.34% for ResNet50, MobileNetV1, and KNN respectively.