Scholarlink

<p>A brain tumor is defined by the abnormal growth of brain cells, some of which may become cancerous. Early detection and treatment of the disease are critical for improving the patients’ quality of life and increasing their lifespan. Artificial intelligence and medical imaging technologies have made significant advances in disease analysis and prediction, particularly in the detection of brain tumors. Extracting relevant features from Magnetic Resonance Imaging (MRI) scans is an important step in the diagnostic process, and several methods have been proposed. Traditional approaches frequently result in treatment delays, which can negatively affect patient outcomes. To address these issues, this study aimed to develop a precise brain tumor detection and classification system using advanced deep learning techniques. Initially, a homomorphic wavelet filter is used during the preprocessing stage to enhance MRI images by minimizing noise and improving image clarity. Subsequently, segmentation is performed using the Fuzzy C-Means (FCM) clustering algorithm combined with the Salp Swarm Algorithm (SSA). SSA’s optimization capabilities of SSA refine the clustering process, resulting in a more accurate delineation of tumor regions. For feature extraction, the ResNet-101 model was employed owing to its deep residual learning framework, which captures complex patterns and features from the segmented images. The classification was carried out using an enhanced GoogLeNet model, which leverages its advanced convolutional architecture to improve tumor detection accuracy by effectively managing extracted features and differentiating between tumor types. Comparative analysis demonstrates that the proposed model outperforms other classifiers, such as SqueezeNet, MobileNetv2, VGG-16, and AlexNet, achieving an accuracy of 98.17%, specificity of 91.34%, and sensitivity of 98.79%.</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

An Efficient Approach for Tumor Grade Classification from MRI Image using Hybrid ResNet-101 with Enhanced GoogLeNet Algorithm

  • 期刊论文

  • 发表日期：

    2025年6月13日

查看全文
  • Kannan Balasubramani,
  • Karthigai Lakshmi Shanmugavel

摘要

A brain tumor is defined by the abnormal growth of brain cells, some of which may become cancerous. Early detection and treatment of the disease are critical for improving the patients’ quality of life and increasing their lifespan. Artificial intelligence and medical imaging technologies have made significant advances in disease analysis and prediction, particularly in the detection of brain tumors. Extracting relevant features from Magnetic Resonance Imaging (MRI) scans is an important step in the diagnostic process, and several methods have been proposed. Traditional approaches frequently result in treatment delays, which can negatively affect patient outcomes. To address these issues, this study aimed to develop a precise brain tumor detection and classification system using advanced deep learning techniques. Initially, a homomorphic wavelet filter is used during the preprocessing stage to enhance MRI images by minimizing noise and improving image clarity. Subsequently, segmentation is performed using the Fuzzy C-Means (FCM) clustering algorithm combined with the Salp Swarm Algorithm (SSA). SSA’s optimization capabilities of SSA refine the clustering process, resulting in a more accurate delineation of tumor regions. For feature extraction, the ResNet-101 model was employed owing to its deep residual learning framework, which captures complex patterns and features from the segmented images. The classification was carried out using an enhanced GoogLeNet model, which leverages its advanced convolutional architecture to improve tumor detection accuracy by effectively managing extracted features and differentiating between tumor types. Comparative analysis demonstrates that the proposed model outperforms other classifiers, such as SqueezeNet, MobileNetv2, VGG-16, and AlexNet, achieving an accuracy of 98.17%, specificity of 91.34%, and sensitivity of 98.79%.