Scholarlink

<p>Lung cancer continues to be one of the leading causes of cancer-related deaths worldwide, largely because early tumors are difficult to identify and the disease often advances before symptoms appear. Improving the accuracy and timing of diagnosis is therefore essential for better treatment outcomes. In this study, a hybrid computer-aided diagnostic approach is presented for classifying lung cancer from histopathological images representing three categories: Adenocarcinoma (ACA), Squamous Cell Carcinoma (SCC), and Benign Lung Tissue (BLT). The process begins with Fuzzy K-Means (FKM) clustering, which is used to clean the images and emphasize the tissue regions relevant for analysis. Following segmentation, features are extracted from six commonly used pre-trained CNN (AlexNet, VGG16, VGG19, ResNet50, ResNet101, and DenseNet201) models. These extracted features are then classified using a Support Vector Machine, where the kernel settings are refined through Bayesian optimization to improve predictive performance. The strength of this approach lies in the combination of FKM-based segmentation, deep feature extraction, and an optimized SVM classifier, all of which contribute to stronger and more consistent results. Among the tested models, ResNet50 achieved the highest accuracy at 93.6%, along with strong sensitivity and specificity. These findings indicate that the proposed method has the potential to serve as a practical and dependable tool in assisting pathologists with lung cancer diagnosis.</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

A hybrid framework integrating Fuzzy K-means segmentation and CNN feature extraction with SVM kernel for lung cancer classification

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Potharla Ramadevi,
  • Raja Das

摘要

Lung cancer continues to be one of the leading causes of cancer-related deaths worldwide, largely because early tumors are difficult to identify and the disease often advances before symptoms appear. Improving the accuracy and timing of diagnosis is therefore essential for better treatment outcomes. In this study, a hybrid computer-aided diagnostic approach is presented for classifying lung cancer from histopathological images representing three categories: Adenocarcinoma (ACA), Squamous Cell Carcinoma (SCC), and Benign Lung Tissue (BLT). The process begins with Fuzzy K-Means (FKM) clustering, which is used to clean the images and emphasize the tissue regions relevant for analysis. Following segmentation, features are extracted from six commonly used pre-trained CNN (AlexNet, VGG16, VGG19, ResNet50, ResNet101, and DenseNet201) models. These extracted features are then classified using a Support Vector Machine, where the kernel settings are refined through Bayesian optimization to improve predictive performance. The strength of this approach lies in the combination of FKM-based segmentation, deep feature extraction, and an optimized SVM classifier, all of which contribute to stronger and more consistent results. Among the tested models, ResNet50 achieved the highest accuracy at 93.6%, along with strong sensitivity and specificity. These findings indicate that the proposed method has the potential to serve as a practical and dependable tool in assisting pathologists with lung cancer diagnosis.