研究目的
To use the Number of Texture Unit as a feature extractor for classification of breast images and compare it with the Gray Level Co-occurrence Matrix, showing that their combination improves classification results.
研究成果
The combination of GLCM and NTU feature extraction techniques improves breast disease classification results, achieving 96.15% AUC. NTU alone performs comparably to GLCM, and future work should focus on optimizing attribute selection and computational efficiency.
研究不足
The paper does not explicitly discuss limitations, but potential areas for optimization include reducing the number of input attributes to minimize computational cost and improving attribute selection strategies.
1：Experimental Design and Method Selection:
The study compares feature extraction techniques (GLCM and NTU) for classifying breast diseases from thermal images using SVM classifier.
2：Sample Selection and Data Sources:
80 images from the Digital Medical Images database, each representing a patient with diagnoses of healthy, benign tumor, or cancer.
3：List of Experimental Equipment and Materials:
Not specified in the paper.
4：Experimental Procedures and Operational Workflow:
Manual ROI segmentation including armpit regions, feature extraction using GLCM (with four directions and six descriptors per direction) and NTU (with eight spectra and six features per spectrum), attribute selection via genetic algorithm, and classification with SVM using RBF kernel.
5：Data Analysis Methods:
Performance evaluation using Area Under the Curve (AUC) and Accuracy (ACC), with comparison to literature results.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
