A computerized global MR image feature analysis scheme to assist diagnosis of breast cancer: a preliminary assessment

Objectives: To develop a new computer-aided detection scheme to compute a global kinetic image feature from the dynamic contrast enhanced breast magnetic resonance imaging (DCE-MRI) and test the feasibility of using the computerized results for assisting classification between the DCE-MRI examinations associated with malignant and benign tumors. Materials and Methods: The scheme registers sequential images acquired from each DCE-MRI examination, segments breast areas on all images, searches for a fraction of voxels that have higher contrast enhancement values and computes an average contrast enhancement value of selected voxels. Combination of the maximum contrast enhancement values computed from two post-contrast series in one of two breasts is applied to predict the likelihood of the examination being positive for breast cancer. The scheme performance was evaluated when applying to a retrospectively collected database including 80 malignant and 50 benign cases. Results: In each of 91% of malignant cases and 66% of benign cases, the average contrast enhancement value computed from the top 0.43% of voxels is higher in the breast depicted suspicious lesions as compared to another negative (lesion-free) breast. In classifying between malignant and benign cases, using the computed image feature achieved an area under a receiver operating characteristic curve of 0.839 with 95% confidence interval of [0.762, 0.898]. Conclusions: We demonstrated that the global contrast enhancement feature of DCE-MRI can be relatively easily and robustly computed without accurate breast tumor detection and segmentation. This global feature provides supplementary information and a higher discriminatory power in assisting diagnosis of breast cancer