Scholarlink

<p>As per World Health Organization (WHO) Lung cancer accounts for 1.8 million fatalities annually. Tobacco smoking is the major cause for around 85% of cases, with other factors being second-hand smoke, air pollution, and specific genetic mutations. Despite the availability of treatments, early detection is crucial yet often hindered by late diagnosis. The study introduces an expressly designed Vision Transformer model, named Lung Cancer Detection Vision Transformer (LCDViT), designed for accurate and reliable detection of lung cancer using histopathological images. The LCDViT model is systematically evaluated using various image and patch sizes, demonstrating superior performance in classifying lung cancer compared to existing deep learning models. The evaluation metrics considered include accuracy, precision, recall, F1-score, confusion matrix, and training vs. validation accuracy and loss curves. The LCDViT model achieves 100% accuracy in training, validation, and testing for an image size of 256 × 256 with a patch size of 16 × 16. Moreover, the integration of Explainable AI technique provides valuable insights into model decision-making, highlighting critical image regions for interpretability and thereby providing reliability and confidence to the practitioner. The LCDViT model demonstrates superior performance compared to existing research on the LC25000 dataset. It achieves higher accuracy and better generalization, setting a new benchmark for lung cancer classification accuracy. This research underscores the potential of LCDViT in clinical applications, offering a robust tool for early lung cancer detection and thereby improving patient outcomes.</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

Enhancing lung cancer diagnostic accuracy and reliability with LCDViT: an expressly developed vision transformer model featuring explainable AI

  • 期刊论文

  • 发表日期：

    2025年1月29日

查看全文
  • Amol Satsangi,
  • K. Srinivas,
  • A. Charan Kumari

摘要

As per World Health Organization (WHO) Lung cancer accounts for 1.8 million fatalities annually. Tobacco smoking is the major cause for around 85% of cases, with other factors being second-hand smoke, air pollution, and specific genetic mutations. Despite the availability of treatments, early detection is crucial yet often hindered by late diagnosis. The study introduces an expressly designed Vision Transformer model, named Lung Cancer Detection Vision Transformer (LCDViT), designed for accurate and reliable detection of lung cancer using histopathological images. The LCDViT model is systematically evaluated using various image and patch sizes, demonstrating superior performance in classifying lung cancer compared to existing deep learning models. The evaluation metrics considered include accuracy, precision, recall, F1-score, confusion matrix, and training vs. validation accuracy and loss curves. The LCDViT model achieves 100% accuracy in training, validation, and testing for an image size of 256 × 256 with a patch size of 16 × 16. Moreover, the integration of Explainable AI technique provides valuable insights into model decision-making, highlighting critical image regions for interpretability and thereby providing reliability and confidence to the practitioner. The LCDViT model demonstrates superior performance compared to existing research on the LC25000 dataset. It achieves higher accuracy and better generalization, setting a new benchmark for lung cancer classification accuracy. This research underscores the potential of LCDViT in clinical applications, offering a robust tool for early lung cancer detection and thereby improving patient outcomes.