Scholarlink

<p>Brain tumour is a grave health issue that affects thousands of individuals globally, either directly or indirectly. A patient’s improved quality of life and the effectiveness of therapy depend on an early and precise identification of brain tumours. In order to detect brain tumours, numerous imaging modalities are employed. The most popular of these methods are MRI and CT scans. Computer-aided analysis of brain pictures has gained popularity recently as a viable method for accurate and reliable brain cancer identification to address the drawbacks of these conventional methods. Hence, in this research, a method for classifying brain tumours from MRI images is created called Precise Brain Tumour Classification (PBTC). Online sources are used to compile the databases. The four stages that the suggested approach is supposed to accomplish initially pre-processing then segmentation after that feature extraction and finally classification phase. Database may contain noise and unnecessary information. Pre-processing is crucial to remove extraneous data and lower noise in brain MRI imaging. Cropping is frequently used to eliminate non-brain areas and lower noise. In this work, a cropping technique is used to calculate the extreme points of the brain region. The two primary morphological processes of dilation and erosion were used to lessen noise. Using Active Contour by Level Set (ACLS), the pre-processing&#xa0;image is routed to the segmentation stage. The feature extraction method employed to extract the features from the segmented image was Gray-Level Co-occurrence matrix (GLCM). Finally, the proposed classifier is based on a Hybrid Recurrent Neural Network-Bidirectional Long Short Memory (HRNN-BiLSTM) structure. The suggested classifier combines Humming Bird Optimization (OHBO) based on opposition and RNN-BiLSTM. In RNN-BiLSTM, the weight variable is chosen by OHBO. The effectiveness of the recommended technique is assessed using MATLAB-implemented metrics such as recall, sensitivity, specificity, accuracy, and precision. The Comparative results show that the proposed model gives the Sensitivity of 94%, Specificity of 96% and Accuracy of 97.8%.</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

Precise brain tumor classification from MRI images with hybrid recurrent neural network-bidirectional LSTM and humming bird optimization

  • 期刊论文

  • 发表日期：

    2025年2月25日

查看全文
  • Gokapay Dilip Kumar,
  • Sachi Nandan Mohanty

摘要

Brain tumour is a grave health issue that affects thousands of individuals globally, either directly or indirectly. A patient’s improved quality of life and the effectiveness of therapy depend on an early and precise identification of brain tumours. In order to detect brain tumours, numerous imaging modalities are employed. The most popular of these methods are MRI and CT scans. Computer-aided analysis of brain pictures has gained popularity recently as a viable method for accurate and reliable brain cancer identification to address the drawbacks of these conventional methods. Hence, in this research, a method for classifying brain tumours from MRI images is created called Precise Brain Tumour Classification (PBTC). Online sources are used to compile the databases. The four stages that the suggested approach is supposed to accomplish initially pre-processing then segmentation after that feature extraction and finally classification phase. Database may contain noise and unnecessary information. Pre-processing is crucial to remove extraneous data and lower noise in brain MRI imaging. Cropping is frequently used to eliminate non-brain areas and lower noise. In this work, a cropping technique is used to calculate the extreme points of the brain region. The two primary morphological processes of dilation and erosion were used to lessen noise. Using Active Contour by Level Set (ACLS), the pre-processing image is routed to the segmentation stage. The feature extraction method employed to extract the features from the segmented image was Gray-Level Co-occurrence matrix (GLCM). Finally, the proposed classifier is based on a Hybrid Recurrent Neural Network-Bidirectional Long Short Memory (HRNN-BiLSTM) structure. The suggested classifier combines Humming Bird Optimization (OHBO) based on opposition and RNN-BiLSTM. In RNN-BiLSTM, the weight variable is chosen by OHBO. The effectiveness of the recommended technique is assessed using MATLAB-implemented metrics such as recall, sensitivity, specificity, accuracy, and precision. The Comparative results show that the proposed model gives the Sensitivity of 94%, Specificity of 96% and Accuracy of 97.8%.