Scholarlink

<p>Worldwide, a high fatality rate in females is caused by common breast cancer. Various imaging techniques were in use to diagnose breast cancer; one of the non-invasive techniques, called thermography, was captured using thermal infrared images, which detect variations of temperature on the skin and can detect smaller tumors in dense breasts that lead to early detection of cancer. It can be used in conjunction with standard imaging modalities. The motivation behind this research is to propose an improved early-stage breast cancer computer-aided diagnostic system that employs a machine learning approach applied to infrared breast thermography thermal images for screening cancer. In the thermal image, the patches show variation under the affected region in abnormal patients compared to normal patient thermograms. This paper proposes an approach to the analysis of thermograms that extracts texture and color features of the breast region and applies an ensemble classifier for classification under the healthy or sick category. Notably, the proposed approach addresses the problem of color thermal images and utilization of thermal temperature matrix. This research utilized the three different thermal image views (Front, Left 90<sup>0</sup>, and Right 90<sup>0</sup>) to train the classifiers with an image augmentation approach on the public dataset DMR-IR database. Various typical machine learning models, for instance, Logistic Regression, Naive Bayes, Artificial Neural Network, Decision tree, K-Nearest Neighbor, Support Vector Machine, and Ensemble of Classifiers, have also been employed for comparative study to diagnose automated breast cancer. The proposed approach results in 97.10% accuracy, a sensitivity of 97.52%, a specificity of 96.73%, and a 0.9940 area under the ROC curve (AUC). The proposed approach by the EoC classifier presented a significant contribution and improvement in classifier accuracy. Hence, the proposed methodology may be suggested to the healthcare industry in conjunction with other imaging modalities that will help in the diagnosis of the early stage of cancer (breast).</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

Ensemble Classification of Breast Cancer Using Texture and Color Features

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Ankur Kumar Aggarwal,
  • Alpana,
  • Mrinal Pandey

摘要

Worldwide, a high fatality rate in females is caused by common breast cancer. Various imaging techniques were in use to diagnose breast cancer; one of the non-invasive techniques, called thermography, was captured using thermal infrared images, which detect variations of temperature on the skin and can detect smaller tumors in dense breasts that lead to early detection of cancer. It can be used in conjunction with standard imaging modalities. The motivation behind this research is to propose an improved early-stage breast cancer computer-aided diagnostic system that employs a machine learning approach applied to infrared breast thermography thermal images for screening cancer. In the thermal image, the patches show variation under the affected region in abnormal patients compared to normal patient thermograms. This paper proposes an approach to the analysis of thermograms that extracts texture and color features of the breast region and applies an ensemble classifier for classification under the healthy or sick category. Notably, the proposed approach addresses the problem of color thermal images and utilization of thermal temperature matrix. This research utilized the three different thermal image views (Front, Left 900, and Right 900) to train the classifiers with an image augmentation approach on the public dataset DMR-IR database. Various typical machine learning models, for instance, Logistic Regression, Naive Bayes, Artificial Neural Network, Decision tree, K-Nearest Neighbor, Support Vector Machine, and Ensemble of Classifiers, have also been employed for comparative study to diagnose automated breast cancer. The proposed approach results in 97.10% accuracy, a sensitivity of 97.52%, a specificity of 96.73%, and a 0.9940 area under the ROC curve (AUC). The proposed approach by the EoC classifier presented a significant contribution and improvement in classifier accuracy. Hence, the proposed methodology may be suggested to the healthcare industry in conjunction with other imaging modalities that will help in the diagnosis of the early stage of cancer (breast).