Scholarlink

<p>As lung cancer (LC) continues to be one of the world’s top causes of cancer-related deaths, precise and sophisticated detection methods are necessary. Traditional deep learning (DL) models typically do not achieve sufficient generalization due to the variability in imaging modality and tumor heterogeneity. This work proposes a novel Hybrid Reflection Cross-Equivariant Network with Quantum Attention to improve LC classification. Rather than just employing the usual convolutional neural networks or construction transformer-based architectures, this work combines Reflection Equivariant Quantum Neural Networks (REQNN) with Cross Attention Mechanism (CAM) performed using improved Pied Kingfisher Optimizer (IPKO) The advantages of this combination are to increase the robustness to some translated transformation in space, increased attention in focus of feature, and faster convergence during training, resulting in improved classification abilities. LC estimation images are taken from three benchmarks: the SEER NSCLC Dataset, the LIDC-LUNA16 CT Scan Dataset, and the LC25000 dataset. Wavelet Adaptive Morphology Perona-Malik. To reduce background noise and eliminate noise, filtering is utilized. Causality-Aware Transformer Networks are then used to segment the data. Feature extraction and classification are done with the Improved Reflection Cross Equivariant Pied Kingfisher Quantum Attention Networks (Imp-RCE-PKQAN) that supports the reliable identification of cancer regions. The experimental study shows that, in comparison to several other sophisticated, cutting-edge models for the detection of lung cancer, Imp-RCE-PKQAN obtains one of the best accuracy rates of 99.9%. With less computational complexity, the suggested model captures morphological structures and spatial dependencies with excellent performance. Unlike existing approaches that address attention or symmetry separately, the proposed method uniquely integrates reflection-equivariant quantum neural networks, cross-attention, and efficient optimization in a single framework, ensuring robust, interpretable, and computationally efficient lung cancer detection.</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 Reflection Cross-Equivariant Quantum Attention Network for Robust Lung Cancer Detection

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Mohammad Yasmeen Kousar,
  • Sadam Vidyasagar Yadav,
  • Erelli Dilip Kumar,
  • Devavarapu Sreenivasarao

摘要

As lung cancer (LC) continues to be one of the world’s top causes of cancer-related deaths, precise and sophisticated detection methods are necessary. Traditional deep learning (DL) models typically do not achieve sufficient generalization due to the variability in imaging modality and tumor heterogeneity. This work proposes a novel Hybrid Reflection Cross-Equivariant Network with Quantum Attention to improve LC classification. Rather than just employing the usual convolutional neural networks or construction transformer-based architectures, this work combines Reflection Equivariant Quantum Neural Networks (REQNN) with Cross Attention Mechanism (CAM) performed using improved Pied Kingfisher Optimizer (IPKO) The advantages of this combination are to increase the robustness to some translated transformation in space, increased attention in focus of feature, and faster convergence during training, resulting in improved classification abilities. LC estimation images are taken from three benchmarks: the SEER NSCLC Dataset, the LIDC-LUNA16 CT Scan Dataset, and the LC25000 dataset. Wavelet Adaptive Morphology Perona-Malik. To reduce background noise and eliminate noise, filtering is utilized. Causality-Aware Transformer Networks are then used to segment the data. Feature extraction and classification are done with the Improved Reflection Cross Equivariant Pied Kingfisher Quantum Attention Networks (Imp-RCE-PKQAN) that supports the reliable identification of cancer regions. The experimental study shows that, in comparison to several other sophisticated, cutting-edge models for the detection of lung cancer, Imp-RCE-PKQAN obtains one of the best accuracy rates of 99.9%. With less computational complexity, the suggested model captures morphological structures and spatial dependencies with excellent performance. Unlike existing approaches that address attention or symmetry separately, the proposed method uniquely integrates reflection-equivariant quantum neural networks, cross-attention, and efficient optimization in a single framework, ensuring robust, interpretable, and computationally efficient lung cancer detection.