Kumar Harish, Taluja Anuradha, Muniyandy Elangovan, Kolli Srinivas
Department of Computer Science and Engineering, SDGI Global University, Ghaziabad, Uttar Pradesh, India.
Department of Computer Science and Engineering, Ajay Kumar Garg Engineering College, Ghaziabad, Uttar Pradesh, India.
Abdom Radiol (NY). 2025 Apr 5. doi: 10.1007/s00261-025-04900-4.
One of the most prevalent cancers in the world is kidney cancer (KC). A precise diagnosis, which is influenced by a number of variables, such as the size or volume of the tumor, the types and stages of the cancer, etc., is essential for the treatment of patients with kidney cancer. In this work two main types of kidney cancer: normal and abnormal, using the accessible KiTS21 dataset of contrast-enhanced CT scans and associated data from patients. Many of these techniques show poor accuracy, which raises doubts regarding their efficiency and dependability. To overcome these limitations, this paper presents the use of a double-decker convolutional neural network with the great wall construction algorithm (DDCNN-GWCA). Hybrid quick conventional bilateral filter improves the quality of pre-processed data by reducing noise while preserving crucial information by using the KiTS21 dataset. Practical Quantum K-Means Clustering is used for segmentation to improve detection efficiency and accuracy. The Q-value Regularized Transformer (QT) is a feature extraction method that combines the power of transformers with Q-value regularization to capture the relevant features. A Double-Decker Convolutional Neural Network's multi-layered architecture is used for classification to identify the classes. The Great Wall Construction Algorithm is an innovative optimization technique that optimizes the hyperparameters of the Double Decker Convolutional Neural Network (DDCNN), ensuring enhanced performance. It obtained scores of 98.9% for the KiTS21 dataset. These results demonstrate the strategy's ability to outperform existing methods and open the way for major advances in the diagnosis of kidney cancer.
肾癌（KC）是世界上最常见的癌症之一。精确诊断受多种因素影响，如肿瘤大小或体积、癌症类型和阶段等，对肾癌患者的治疗至关重要。在这项工作中，利用可获取的KiTS21增强CT扫描数据集以及患者的相关数据，对两种主要类型的肾癌：正常和异常进行研究。许多这些技术显示出较差的准确性，这引发了对其效率和可靠性的质疑。为克服这些限制，本文提出使用带有长城构建算法（DDCNN - GWCA）的双层卷积神经网络。混合快速传统双边滤波器通过使用KiTS21数据集减少噪声同时保留关键信息，提高了预处理数据的质量。实用量子K均值聚类用于分割以提高检测效率和准确性。Q值正则化变换器（QT）是一种特征提取方法，它将变换器的能力与Q值正则化相结合以捕获相关特征。双层卷积神经网络的多层架构用于分类以识别类别。长城构建算法是一种创新的优化技术，用于优化双层卷积神经网络（DDCNN）的超参数，确保性能提升。它在KiTS21数据集上获得了98.9%的分数。这些结果证明了该策略优于现有方法的能力，并为肾癌诊断的重大进展开辟了道路。
