Scholarlink

<p>Brain tumor classification has become a critical entry point in enhancing the diagnosis and the subsequent management in medical imaging. In an attempt to design an efficient classifier for brain tumors, this study introduces the hybrid quantum-classical neural networks (H-QNNs). The model comprises standard convolutional layers for the extraction of relevant features from input images with a quantum neural network based on ZZFeatureMap with RealAmplitudes Ansatz for enhanced feature identification. MRI images of four tumor classes were assessed in the study, which include glioma, meningioma, pituitary tumor, and no tumor. The proposed framework led to a classification accuracy of 95.58% with precision, recall, and F1-score at 95.58%, significantly surpassing legacy methods. Data imbalance was resolved using data augmentation and weighted cross-entropy loss, while methods such as weight decay were used to increase model resilience. Findings indicate the future application of QML in improving the efficiency of computer-aided image analysis for medical diagnosis and accuracy of classification. Specific strategies mentioned for future work include extending the method to larger datasets and improving quantum components to support additional tasks.</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 tumor diagnosis using hybrid quantum-classical neural networks

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Deepak Ranga,
  • Sunil Prajapat,
  • Kranti Kumar,
  • Pankaj Kumar

摘要

Brain tumor classification has become a critical entry point in enhancing the diagnosis and the subsequent management in medical imaging. In an attempt to design an efficient classifier for brain tumors, this study introduces the hybrid quantum-classical neural networks (H-QNNs). The model comprises standard convolutional layers for the extraction of relevant features from input images with a quantum neural network based on ZZFeatureMap with RealAmplitudes Ansatz for enhanced feature identification. MRI images of four tumor classes were assessed in the study, which include glioma, meningioma, pituitary tumor, and no tumor. The proposed framework led to a classification accuracy of 95.58% with precision, recall, and F1-score at 95.58%, significantly surpassing legacy methods. Data imbalance was resolved using data augmentation and weighted cross-entropy loss, while methods such as weight decay were used to increase model resilience. Findings indicate the future application of QML in improving the efficiency of computer-aided image analysis for medical diagnosis and accuracy of classification. Specific strategies mentioned for future work include extending the method to larger datasets and improving quantum components to support additional tasks.