Scholarlink

<p>One of the most prevalent causes of mortality worldwide is lung cancer. It is imperative to improve patient outcomes by early detection and accurate classification of pulmonary nodules. In this paper, we have proposed a framework for the early and accurate detection of lung cancer. The proposed model employs sophisticated deep learning and image processing techniques to resolve the challenge of detecting micro nodule and distinguishing between malignant and benign pulmonary nodules. With the LIDC-IDRI dataset, we pre-processed CT images by employing median filtering for noise reduction and thresholding techniques for nodule segmentation. Thresholding is implemented in connected component analysis to enable the identification of even the smallest potential nodules, thereby increasing the probability of early malignant nodule detection. The feature extraction process was conducted using pre-trained transfer learning models, including ResNet50, VGG16, and EfficientNet-B0. Subsequent to data normalization and feature selection, the Random Forest classifier was employed. The proposed framework, EVRNet_RF model, integrates robust pre-processing techniques, advanced feature extraction using transfer learning models, and precise feature selection through PCA. Notably, EVRNet_RF has demonstrated superior performance over other models with 95% accuracy. </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

Optimizing lung nodule classification through the integration of image processing and transfer learning techniques

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Takreem Fatima Khan,
  • Swaleha Zubair

摘要

One of the most prevalent causes of mortality worldwide is lung cancer. It is imperative to improve patient outcomes by early detection and accurate classification of pulmonary nodules. In this paper, we have proposed a framework for the early and accurate detection of lung cancer. The proposed model employs sophisticated deep learning and image processing techniques to resolve the challenge of detecting micro nodule and distinguishing between malignant and benign pulmonary nodules. With the LIDC-IDRI dataset, we pre-processed CT images by employing median filtering for noise reduction and thresholding techniques for nodule segmentation. Thresholding is implemented in connected component analysis to enable the identification of even the smallest potential nodules, thereby increasing the probability of early malignant nodule detection. The feature extraction process was conducted using pre-trained transfer learning models, including ResNet50, VGG16, and EfficientNet-B0. Subsequent to data normalization and feature selection, the Random Forest classifier was employed. The proposed framework, EVRNet_RF model, integrates robust pre-processing techniques, advanced feature extraction using transfer learning models, and precise feature selection through PCA. Notably, EVRNet_RF has demonstrated superior performance over other models with 95% accuracy.