Abstract #2927

Prediction of Malignant Breast Lesions from MRI Features: A Comparison of Artificial Neural Network and Logistic Regression Techniques

Ke Nie¹, Christine E. McLaren², Wen-Pin Chen², Jeon-Hor Chen^1,3, Orhan Nalcioglu¹, Min-Ying Lydia Su¹

¹Tu & Yuen Center for Functional Onco-Imaging, University of California, Irvine, Irvine, CA, USA; ²Department of Epidemiology, University of California, Irvine, Irvine, CA, USA; ³Department of Radiology, China Medical University, Taichung, Taiwan

This study compared two different approaches using artificial neural network (ANN) and logistic regression analysis (LRA) for selecting diagnostic models to differentiate between malignant (N=43) and benign (N=28) lesions. For each case, 8 parameters were used to characterize morphology, 10 GCLM and 14 Laws features were used to characterize enhancement texture (homogeneity) within the lesion. The diagnostic performance of classifiers obtained by ANN or different LRA models was similar. While ANN is more robust and does not require a high level of operator judgment, LRA may be used complimentarily to understand the role of the selected predictors.