Scholarlink

<p>Lung cancer remains one of the most prevalent and fatal malignancies worldwide, underscoring the critical need for accurate, efficient, and interpretable diagnostic systems. This study introduces a novel computational framework for the automated classification of lung histopathology images into three clinically relevant categories: benign tissue, adenocarcinoma, and squamous cell carcinoma. The proposed method leverages deep features extracted from a pre-trained VGG16 network to capture subtle morphological patterns indicative of malignancy. To address feature redundancy and enhance both diagnostic precision and computational efficiency, Grey Wolf Optimization (GWO) is employed to reduce the high-dimensional feature space from 8192 to 3424 informative attributes. These optimized features are subsequently used to train a custom Feedforward Neural Network (FFNN), which achieves a state-of-the-art test accuracy of 99.74% on the LC25000 dataset comprising 15,000 annotated histopathological images. In comparison, a baseline model trained on the complete feature set attained a slightly lower accuracy of 99.61%, while incurring substantially higher computational costs. Furthermore, SHAP-based explainability is integrated into the framework, providing transparent insights into feature contributions and enhancing the interpretability of the model’s decision-making process. The results demonstrate that the proposed approach not only achieves superior classification accuracy but also reduces computational overhead, thereby offering a practical solution for real-world clinical deployment. By uniting efficiency with interpretability, this work aims to support pathologists in timely and reliable lung cancer diagnosis, potentially contributing to improved patient outcomes through earlier clinical intervention.<!--Query ID="Q1" Text="Please confirm if the author names are presented accurately and in the correct sequence (given name, middle nameinitial, family name). Author 1 Given name: [Koushlendra Kumar] Last name [Singh], Author 2 Given name: [Pramod Kumar] Last name [Soni]. Also, kindly confirm the details in the metadata are correct." Resolved="yes"--></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

Feature-optimized deep neural network with explainable AI for lung cancer histopathology classification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月4日

查看全文
  • Onkar Singh,
  • Koushlendra Kumar Singh,
  • Chandrasen Pandey,
  • Pramod Kumar Soni

摘要

Lung cancer remains one of the most prevalent and fatal malignancies worldwide, underscoring the critical need for accurate, efficient, and interpretable diagnostic systems. This study introduces a novel computational framework for the automated classification of lung histopathology images into three clinically relevant categories: benign tissue, adenocarcinoma, and squamous cell carcinoma. The proposed method leverages deep features extracted from a pre-trained VGG16 network to capture subtle morphological patterns indicative of malignancy. To address feature redundancy and enhance both diagnostic precision and computational efficiency, Grey Wolf Optimization (GWO) is employed to reduce the high-dimensional feature space from 8192 to 3424 informative attributes. These optimized features are subsequently used to train a custom Feedforward Neural Network (FFNN), which achieves a state-of-the-art test accuracy of 99.74% on the LC25000 dataset comprising 15,000 annotated histopathological images. In comparison, a baseline model trained on the complete feature set attained a slightly lower accuracy of 99.61%, while incurring substantially higher computational costs. Furthermore, SHAP-based explainability is integrated into the framework, providing transparent insights into feature contributions and enhancing the interpretability of the model’s decision-making process. The results demonstrate that the proposed approach not only achieves superior classification accuracy but also reduces computational overhead, thereby offering a practical solution for real-world clinical deployment. By uniting efficiency with interpretability, this work aims to support pathologists in timely and reliable lung cancer diagnosis, potentially contributing to improved patient outcomes through earlier clinical intervention.