Scholarlink

With its increasing incidence rates, skin cancer is a global health concern that necessitates innovative strategies for early identification and better patient outcomes. In order to meet the critical need for early diagnosis in a variety of healthcare contexts, this work presents LesionLensPro, an innovative ensemble deep learning model for skin cancer detection. By combining the EfficientNetV2 and ResNet architectures, the model achieves outstanding results in skin lesion classification, with a ROC-AUC of 94.169%, accuracy of 98.929%, and F1-Score of 96.092%. LesionLensPro uses a SLICE-3D dataset of histopathologically confirmed lesions from multiple continents to achieve remarkable results across a variety of populations and skin types. The model’s ability to process both picture data and metadata information is a significant advancement in AI-driven medical imaging. Tested on the dataset, which simulates real-world clinical scenarios, LesionLensPro shows promise for practical healthcare applications, particularly in underserved populations. This research aims to accelerate progress in AI-assisted skin cancer screening, potentially improving early detection rates and patient outcomes worldwide.

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

LesionLensPro: Skin Cancer Detection Using an Ensemble Approach with EfficientNetV2 and ResNet

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Deepan Adak,
  • Gaurav Verma

摘要

With its increasing incidence rates, skin cancer is a global health concern that necessitates innovative strategies for early identification and better patient outcomes. In order to meet the critical need for early diagnosis in a variety of healthcare contexts, this work presents LesionLensPro, an innovative ensemble deep learning model for skin cancer detection. By combining the EfficientNetV2 and ResNet architectures, the model achieves outstanding results in skin lesion classification, with a ROC-AUC of 94.169%, accuracy of 98.929%, and F1-Score of 96.092%. LesionLensPro uses a SLICE-3D dataset of histopathologically confirmed lesions from multiple continents to achieve remarkable results across a variety of populations and skin types. The model’s ability to process both picture data and metadata information is a significant advancement in AI-driven medical imaging. Tested on the dataset, which simulates real-world clinical scenarios, LesionLensPro shows promise for practical healthcare applications, particularly in underserved populations. This research aims to accelerate progress in AI-assisted skin cancer screening, potentially improving early detection rates and patient outcomes worldwide.