Scholarlink

<p>Skin cancer is widely known as the most frequent cancer type worldwide. Early cancer detection is critical as it might cause death if not diagnosed in the early stages. It can be identified through open eyes, but due to great inter-class similarities and intra-class variances, it is difficult to notice. Because of the global prevalence of skin cancers, deep learning is used to create various automated methods to assist physicians in early skin lesion diagnosis. The primary goal of this study is to propose a new tested and calibrated model of deep learning for multiple class categorization of skin lesions. The research community developed numerous Computer-Aided Diagnostic systems (CADs) to support lesion detection. However, this study proposes an automated system of CAD designed for multi-class classification of skin lesions. The dataset used by the authors is HAM1000, and a comprehensive study was performed on three pre-trained Convolutional Neural Network (CNN) models and three hybrid CNN models. To increase efficacy and performance, the Model is suggested and it is multilayered that is designed with great care having multiple with varied sizes of filter, but fewer parameters too. The models’ effectiveness was tested on parameters like precision, the phenomenon of recall, and the score of F1. Maximum accuracy is 91.63% which is reported through a model which is hybrid, and it is proposed in this paper. </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 deep CNN model for multi-class classification of skin lesion

  • 期刊论文

  • 发表日期：

    2025年7月9日

查看全文
  • Mohammad Ahmar Khan,
  • Deependra Rastogi,
  • Prashant Johri,
  • Ahmad Al-Taani,
  • Vishwadeepak Singh Baghela,
  • Kumud

摘要

Skin cancer is widely known as the most frequent cancer type worldwide. Early cancer detection is critical as it might cause death if not diagnosed in the early stages. It can be identified through open eyes, but due to great inter-class similarities and intra-class variances, it is difficult to notice. Because of the global prevalence of skin cancers, deep learning is used to create various automated methods to assist physicians in early skin lesion diagnosis. The primary goal of this study is to propose a new tested and calibrated model of deep learning for multiple class categorization of skin lesions. The research community developed numerous Computer-Aided Diagnostic systems (CADs) to support lesion detection. However, this study proposes an automated system of CAD designed for multi-class classification of skin lesions. The dataset used by the authors is HAM1000, and a comprehensive study was performed on three pre-trained Convolutional Neural Network (CNN) models and three hybrid CNN models. To increase efficacy and performance, the Model is suggested and it is multilayered that is designed with great care having multiple with varied sizes of filter, but fewer parameters too. The models’ effectiveness was tested on parameters like precision, the phenomenon of recall, and the score of F1. Maximum accuracy is 91.63% which is reported through a model which is hybrid, and it is proposed in this paper.