Functional Diversity applied to the false positive reduction in breast tissues based on digital mammography

Breast cancer is currently the most common in female patients and the second with the highest mortality rate. The primary responsibility for these alarming statistical data, which has been growing in recent years, are still factors of external risks such as excessive consumption of alcohol, tobacco, processed foods, sedentary lifestyle, obesity or any item associated with an unbalanced lifestyle. Also, another major impact factor is related to late diagnosis and treatment. With this, several mechanisms, such as CAD systems, are being developed to assist specialists in rapid and early diagnosis. This work proposes an approach to reduce false positives. To evaluate and validate the proposed methodology regions extracted from the DDSM database using a CAD system were used. In the proposed methodology used texture descriptors based on functional diversity indexes for the extraction of characteristics, followed by the classification of regions of interest in mass and non-mass. The results were promising, reaching rates of accuracy, sensitivity, specificity, kappa index and area under the ROC curve of 92.29%, 90.15%, 95.65%, 0.841 and 0.939, respectively.