Haralick Textures Features as an Identifier of Triple Negative Breast Cancer

This study aims to discern and analyze statistical differences in Haralick textures features in mammography images, specifically comparing patients diagnosed with triple-negative breast cancer with those with benign tumors. Methods: This study included two data sets (triple-negative breast cancer and benign tumor), each with 29 mammography images. Haralick textures features were extracted using MATLAB software (MathWorks, Natick, MA, USA). The Kolmogorov-Smirnov test was applied to verify the normality of the data distribution. Given that the data did not follow a normal distribution, the nonparametric Kruskal-Wallis test was used in the statistical analysis. Results: The texture characteristics that showed significant discriminatory power between the analyzed groups were: Contrast, Entropy, Sum of Entropy, Difference Entropy, and Informational Correlation Measure 2. Conclusion: The findings of this study support the adoption of a broad and iterative approach to texture analysis, involving the integration of relevant descriptors with complementary analytical techniques, aiming to improve the ability to differentiate between distinct clinical conditions.