Logistic regression model for breast cancer automatic diagnosis

Breast cancer is the uncontrolled growth of breast cells. It represents the second cause of cancer death in women worldwide. It is important for patients to understand their disease and know what to expect in the future so that they can make decisions about treatment, rehabilitation, financial aid decisions and personal matters. This paper presents an approach for diagnosing breast cancer based on a set of input variables that describe some characteristics of tumor images. The proposed approach builds a binary logistic model that classifies between malignant and benign cases. The approach is applied to the Wisconsin Diagnostic Breast Cancer (WDBC) dataset. Experimental results show that the regression model that is statistically significant includes only the area, texture, concavity and symmetry features of a tumor. In addition to the simplicity of the used model, the reduced set of features gives performance measures that outperform similar approaches. Accordingly, the presented approach can be used for feature selection and reduction of the breast cancer data.