Scholarlink

Skin diseases globally present a significant health challenge, impacting individuals physically and emotionally while contributing to societal stigmas. The economic burden is also intensified by inaccurate diagnoses, highlighting the urgent need for efficient and accessible diagnostic tools to alleviate both the financial and personal toll of skin diseases. This research investigates the application of Vision Transformers (ViTs) in dermatological classification, employing a diverse dataset sourced from The International Skin Imaging Collaboration (ISIC) and enriched with examples from HAM10000. The study develops a web application for precise dermatological image classification, achieving an accuracy of 90.41% and an F1 score of 90.39%. The user-friendly web application allows healthcare professionals and individuals to upload images for automated analysis, facilitating early disease identification and intervention. This approach contributes in improving diagnostic tools for dermatological conditions, showcasing the adaptability of deep learning frameworks in medical image analysis for future healthcare advancements.

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

Advanced Diagnostic Framework with Vision Transformer for Multiclass Skin Disease Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Barunaditya Mohanty,
  • Ayush Singhal,
  • Brajesh Kumar,
  • Praveen Kumar Yadav,
  • Priyadarshi Kanungo,
  • Ram Chandra Barik

摘要

Skin diseases globally present a significant health challenge, impacting individuals physically and emotionally while contributing to societal stigmas. The economic burden is also intensified by inaccurate diagnoses, highlighting the urgent need for efficient and accessible diagnostic tools to alleviate both the financial and personal toll of skin diseases. This research investigates the application of Vision Transformers (ViTs) in dermatological classification, employing a diverse dataset sourced from The International Skin Imaging Collaboration (ISIC) and enriched with examples from HAM10000. The study develops a web application for precise dermatological image classification, achieving an accuracy of 90.41% and an F1 score of 90.39%. The user-friendly web application allows healthcare professionals and individuals to upload images for automated analysis, facilitating early disease identification and intervention. This approach contributes in improving diagnostic tools for dermatological conditions, showcasing the adaptability of deep learning frameworks in medical image analysis for future healthcare advancements.