Improving cancer detection through computer-aided diagnosis: A comprehensive analysis of nonlinear and texture features in breast thermograms.

Breast cancer is a significant health issue for women, characterized by its high rates of mortality and sickness. However, its early detection is crucial for improving patient outcomes. Thermography, which measures temperature variations between healthy and cancerous tissues, offers a promising approach for early diagnosis. This study proposes a novel method for analyzing breast thermograms. The method segments suspicious masses, extracts relevant features, and classifies them as benign or malignant. While the chaotic indices, including Lyapunov Exponent (LE), Fractal Dimension (FD), Kolmogorov-Sinai Entropy (KSE), and Correlation Dimension (CD), are employed for nonlinear analysis, the Gray-Level Co-occurrence Matrix (GLCM) method utilized for extracting the texture features. The effectiveness of the proposed approach is enhanced by integrating texture and complexity features. Besides, to optimize feature selection and reduce redundancy, a metaheuristic optimization technique called Non-Dominated Sorting Genetic Algorithm (NSGA III) is applied. The proposed method utilizes various machine learning algorithms, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Pattern recognition Network (Pat net), and Fitting neural Network (Fit net), for classification. ten-fold cross-validation ensures robust performance evaluation. The achieved accuracy of 98.65%, emphasizes the superior performance of the proposed method in thermograms breast cancer diagnosis.