Scholarlink

<p>Cancer remains the second leading cause of death worldwide, with brain tumors contributing significantly to cancer-related fatalities. Early and accurate diagnosis of brain tumors is crucial for timely therapeutic intervention and improved patient outcomes. Recent advancements in medical imaging, particularly in image classification, have notably enhanced the efficacy of computer-aided diagnostic systems. This paper proposes five hybrid models: VGG16 + LSTM + SVM, VGG19 + LSTM + SVM, ResNet50 + LSTM + SVM, ResNet152 + LSTM + SVM, and EfficientNetB3 + LSTM + SVM, for the multi-class classification of brain tumors. These models combine the feature extraction power of convolutional neural networks (VGG16, VGG19, ResNet50, ResNet152, and EfficientNetB3) with the sequential learning capabilities of Long Short-Term Memory (LSTM) networks and the robust classification strength of Support Vector Machines (SVM). The methodology begins with data augmentation to address class imbalance and increase the diversity of the training dataset. Features extracted by the convolutional layers are fed into LSTM networks, which capture temporal dependencies, and the resulting outputs are classified using SVM. Experimental results demonstrate that the proposed models outperform current state-of-the-art approaches, with the ResNet152 + LSTM + SVM model achieving 99% accuracy, 99% precision, 99% recall, and 99% F1 score using the RMSprop optimizer.</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 CNN + LSTM + SVM model for advanced brain tumor classification

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Moriba Bérété,
  • Amira Echtioui,
  • Lamia Sellami,
  • Ahmed AlMokhtar Ben Hmida

摘要

Cancer remains the second leading cause of death worldwide, with brain tumors contributing significantly to cancer-related fatalities. Early and accurate diagnosis of brain tumors is crucial for timely therapeutic intervention and improved patient outcomes. Recent advancements in medical imaging, particularly in image classification, have notably enhanced the efficacy of computer-aided diagnostic systems. This paper proposes five hybrid models: VGG16 + LSTM + SVM, VGG19 + LSTM + SVM, ResNet50 + LSTM + SVM, ResNet152 + LSTM + SVM, and EfficientNetB3 + LSTM + SVM, for the multi-class classification of brain tumors. These models combine the feature extraction power of convolutional neural networks (VGG16, VGG19, ResNet50, ResNet152, and EfficientNetB3) with the sequential learning capabilities of Long Short-Term Memory (LSTM) networks and the robust classification strength of Support Vector Machines (SVM). The methodology begins with data augmentation to address class imbalance and increase the diversity of the training dataset. Features extracted by the convolutional layers are fed into LSTM networks, which capture temporal dependencies, and the resulting outputs are classified using SVM. Experimental results demonstrate that the proposed models outperform current state-of-the-art approaches, with the ResNet152 + LSTM + SVM model achieving 99% accuracy, 99% precision, 99% recall, and 99% F1 score using the RMSprop optimizer.