Scholarlink

<p>Skin cancer constitutes a third of all cancer diagnoses worldwide, with its incidence steadily increasing over recent decades. The introduction of dermoscopy has significantly improved the diagnostic accuracy for skin cancer. However, dermatologists face considerable challenges in accurately diagnosing skin cancer due to the similar appearances of different types. The diagnostic accuracy among dermatologist’s ranges between 62 and 80%. The research community has achieved significant progress in creating automated systems to aid dermatologists in decision-making. This work presents a very precise automated computer-aided diagnostic (CAD) system for multi-class skin cancer classification. Our suggested strategy improves the accuracy of both skilled dermatologist and existing deep learning techniques in MCS cancer classification. Utilizing the HAM10000 dataset, we fine-tuned models across seven skin cancer classes and conducted a comparative study to evaluate the performance of three pre-trained convolutional neural networks (CNNs) and three ensemble models. Our results demonstrated a maximum accuracy of 92.16% for the best-performing individual model and an impressive 94.17% accuracy for the ensemble model. We suggest for taking advantage of InceptionV3, which for MCS cancer classification owing to its improved architecture and enhanced accuracy. This research underscores the capability of sophisticated deep learning models to improve the diagnostic process for skin cancer, providing a significant resource to assist dermatologists in attaining more precise diagnoses.</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

Enhanced Skin Cancer Diagnosis via Deep Convolutional Neural Networks with Ensemble Learning

  • 期刊论文

  • 发表日期：

    2025年1月30日

查看全文
  • Mohd Anas Khan,
  • Shahzad Alam,
  • Waseem Ahmed

摘要

Skin cancer constitutes a third of all cancer diagnoses worldwide, with its incidence steadily increasing over recent decades. The introduction of dermoscopy has significantly improved the diagnostic accuracy for skin cancer. However, dermatologists face considerable challenges in accurately diagnosing skin cancer due to the similar appearances of different types. The diagnostic accuracy among dermatologist’s ranges between 62 and 80%. The research community has achieved significant progress in creating automated systems to aid dermatologists in decision-making. This work presents a very precise automated computer-aided diagnostic (CAD) system for multi-class skin cancer classification. Our suggested strategy improves the accuracy of both skilled dermatologist and existing deep learning techniques in MCS cancer classification. Utilizing the HAM10000 dataset, we fine-tuned models across seven skin cancer classes and conducted a comparative study to evaluate the performance of three pre-trained convolutional neural networks (CNNs) and three ensemble models. Our results demonstrated a maximum accuracy of 92.16% for the best-performing individual model and an impressive 94.17% accuracy for the ensemble model. We suggest for taking advantage of InceptionV3, which for MCS cancer classification owing to its improved architecture and enhanced accuracy. This research underscores the capability of sophisticated deep learning models to improve the diagnostic process for skin cancer, providing a significant resource to assist dermatologists in attaining more precise diagnoses.