Logistic Regression Analysis of Breast Cancer From Mammographic Evaluation

Logistic regression analysis is used to differentiate malignant from benign in a group of patients with proved breast lesions on the base of morphological data extracted from the conventional mammogram. Our database include 122 patients' records consisting 12 qualitative variables. The database is randomly divided into the training and validation samples including 82 and 40 patients' records respectively. The training and validation samples are used to construct the logistic regression modelas a classifier and to validate its performance respectively. Finally, important criteria such as sensitivity, specificity, accuracy and receiver operating characteristic curve (ROC) analysis for this method as well as that of the radiologist are compared. Our results show that the logistic regression model is able to classify correctly 31 out of 40 cases presented in the validation sample. Comparing the output of this method with that of the radiologist shows a reasonable diagnostic accuracy 78%, a high specificity (81%) and a moderate sensitivity (72%).

Keywords:

Breast cancer , Logistic regression analysis , ROC curve