Scholarlink

<p>Pancreatic cancer is the most aggressive cancer with a high mortality rate, and hence, early detection is crucial to improve patient outcomes. Numerous Deep Learning methods have been established for enhancing pancreatic cancer detection in recent years. Despite the promising solution in extracting the spatial features, the existing Convolutional Neural Network (CNN) models struggle to capture the global features, which leads to performance degradation issues. Hence, this research proposes the Hybrid Bottleneck Efficient attention-based Quantum rectified linear unit applied lightweight convolutional neural network (HBoEQN) model to address the limitations in the convolutional neural networks. Specifically, the proposed model combines the Hybrid Bottleneck efficient Attention (HBoEA) module to capture the global correlation of input features, resulting in reducing the errors and contributing to accurate detection. Besides, the Interactive Leadership Optimization (InLOpt) algorithm optimally tunes the HBoEQN model’s parameters, thus enabling accurate performance. In line with this, the research exploits the Hybrid Bottleneck Efficient Attention-based U-Net (HBoEA-Net) model for effectively delineating the tumor regions, which mitigates the computational complexity and eliminates over-fitting issues. Extensive experiments demonstrate that the HBoEQN model using sophisticated mechanisms achieves 97.26% of accuracy, 98.08% of sensitivity, 97.87% F1 score, 98.08% recall, and 97.67% precision, and 96.45% specificity, by outperforming the other state-of-the-art methods using the Decathlon dataset. Segmentation results show that the HBoEA-Net module achieves a Dice Coefficient of 1.93, a Hausdorff Distance of 0.98, an Intersection over Union (IOU) of 0.98, and sensitivity and specificity of 0.98, showing its effective segmentation performance. Moreover, the proposed HBoEQN method improves the feature representation, reduces the computation complexity, and enhances the overall precision, contributing to potential applications for pancreatic 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

HBoEQN: Hybrid Attention-Based Quantum ReLU-Applied Lightweight Deep Learning Model for Pancreatic Cancer Detection Using Computed Tomography

  • 期刊论文

  • 发表日期：

    2025年11月10日

查看全文
  • Suraj Vinod Dhole,
  • Amit Kamalakar Gaikwad

摘要

Pancreatic cancer is the most aggressive cancer with a high mortality rate, and hence, early detection is crucial to improve patient outcomes. Numerous Deep Learning methods have been established for enhancing pancreatic cancer detection in recent years. Despite the promising solution in extracting the spatial features, the existing Convolutional Neural Network (CNN) models struggle to capture the global features, which leads to performance degradation issues. Hence, this research proposes the Hybrid Bottleneck Efficient attention-based Quantum rectified linear unit applied lightweight convolutional neural network (HBoEQN) model to address the limitations in the convolutional neural networks. Specifically, the proposed model combines the Hybrid Bottleneck efficient Attention (HBoEA) module to capture the global correlation of input features, resulting in reducing the errors and contributing to accurate detection. Besides, the Interactive Leadership Optimization (InLOpt) algorithm optimally tunes the HBoEQN model’s parameters, thus enabling accurate performance. In line with this, the research exploits the Hybrid Bottleneck Efficient Attention-based U-Net (HBoEA-Net) model for effectively delineating the tumor regions, which mitigates the computational complexity and eliminates over-fitting issues. Extensive experiments demonstrate that the HBoEQN model using sophisticated mechanisms achieves 97.26% of accuracy, 98.08% of sensitivity, 97.87% F1 score, 98.08% recall, and 97.67% precision, and 96.45% specificity, by outperforming the other state-of-the-art methods using the Decathlon dataset. Segmentation results show that the HBoEA-Net module achieves a Dice Coefficient of 1.93, a Hausdorff Distance of 0.98, an Intersection over Union (IOU) of 0.98, and sensitivity and specificity of 0.98, showing its effective segmentation performance. Moreover, the proposed HBoEQN method improves the feature representation, reduces the computation complexity, and enhances the overall precision, contributing to potential applications for pancreatic cancer detection.