Scholarlink

<p>Effective detection of skin cancer utilizing images is a challenging issue in the healthcare domain. Skin disease detection is a time-consuming operation and if detected at later stages, may lead to death. This paper devises a novel hybrid Quantum Dialated Convolution Fused Neural Network (QDCFN-Net) for skin cancer detection. Here, the skin image obtained from the dataset is considered as input and is preprocessed by the Laplacian filter. After that, skin lesions are isolated from the background effectively by the LadderNet and afterwards, data augmentation is performed by geometric transformation, mixing images pixel averaging (mixup), colorspace transformation, and overlaying crops (CutMix). Consequently, feature extraction is accomplished, wherein features like Haralick texture features, statistical features, Complete Local Binary Pattern (CLBP), Shape Local Binary Texture (SLBT), and Fuzzy Local Binary Pattern (FLBP) are mined. Then, skin cancer detection is executed based on QDCFN-Net, which is obtained by the amalgamation of Deep Neural Network (DNN) and Quantum Dilated Convolutional Neural Network (QDCNN). Furthermore, the estimation of QDCFN-Net is assessed depending on False Positive Rate (FPR), accuracy, True Positive Rate (TPR), and specificity and the best values obtained by the QDCFN-Net are 9.63, 89.68, 90.37, and 90.37%, respectively.</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

Feature extraction and hybrid DNN-QDCNN for skin cancer detection

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • G. Nalinipriya,
  • M. Arun,
  • G. Vasavi,
  • Sreenu Ponnada

摘要

Effective detection of skin cancer utilizing images is a challenging issue in the healthcare domain. Skin disease detection is a time-consuming operation and if detected at later stages, may lead to death. This paper devises a novel hybrid Quantum Dialated Convolution Fused Neural Network (QDCFN-Net) for skin cancer detection. Here, the skin image obtained from the dataset is considered as input and is preprocessed by the Laplacian filter. After that, skin lesions are isolated from the background effectively by the LadderNet and afterwards, data augmentation is performed by geometric transformation, mixing images pixel averaging (mixup), colorspace transformation, and overlaying crops (CutMix). Consequently, feature extraction is accomplished, wherein features like Haralick texture features, statistical features, Complete Local Binary Pattern (CLBP), Shape Local Binary Texture (SLBT), and Fuzzy Local Binary Pattern (FLBP) are mined. Then, skin cancer detection is executed based on QDCFN-Net, which is obtained by the amalgamation of Deep Neural Network (DNN) and Quantum Dilated Convolutional Neural Network (QDCNN). Furthermore, the estimation of QDCFN-Net is assessed depending on False Positive Rate (FPR), accuracy, True Positive Rate (TPR), and specificity and the best values obtained by the QDCFN-Net are 9.63, 89.68, 90.37, and 90.37%, respectively.