Scholarlink

<p>Brain tumor classification is crucial for effective patient care, but traditional MRI-based methods often face accuracy limitations, especially in distinguishing between tumor types. This study introduces a novel Quantum Convolutional Neural Network (QCNN) architecture that leverages quantum embedding, sparse input indexing, and four-qubit quantum convolution layers to enhance classification accuracy and efficiency. Developed using PennyLane and TensorFlow Quantum, our QCNN achieved a testing accuracy of 92.13% on a dataset of over 3000 MRI scans, matching the performance of the classical ResNet50 model while reducing training time from 64.95&#xa0;s to just 1.1&#xa0;s. These results suggest that QCNNs offer a promising new approach for improving brain tumor diagnostics, with the potential for faster and more accurate real-time medical applications, despite challenges in hardware limitations and model interpretability.</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

Advancing neuroimaging with quantum convolutional neural networks for brain tumor detection

  • 期刊论文

  • 发表日期：

    2025年1月27日

查看全文
  • Amrita Ticku,
  • Vaibhav Sangwan,
  • Sanket Balani,
  • Sriti Jha,
  • Sahil Rawat,
  • Anu Rathee,
  • Deepika Yadav

摘要

Brain tumor classification is crucial for effective patient care, but traditional MRI-based methods often face accuracy limitations, especially in distinguishing between tumor types. This study introduces a novel Quantum Convolutional Neural Network (QCNN) architecture that leverages quantum embedding, sparse input indexing, and four-qubit quantum convolution layers to enhance classification accuracy and efficiency. Developed using PennyLane and TensorFlow Quantum, our QCNN achieved a testing accuracy of 92.13% on a dataset of over 3000 MRI scans, matching the performance of the classical ResNet50 model while reducing training time from 64.95 s to just 1.1 s. These results suggest that QCNNs offer a promising new approach for improving brain tumor diagnostics, with the potential for faster and more accurate real-time medical applications, despite challenges in hardware limitations and model interpretability.