Optimizing both feature selection and training classifiers for mass and nonmass lesions improves the accuracy of computer-aided diagnosis (CAD) for breast MR imaging, according to a study published in Radiology.

Researchers from Canada sought to determine suitable features and optimal classifier design for a CAD system to help clinicians differentiate between mass and nonmass enhancements detected by contrast MRI breast imaging.

The researchers respectively collected 280 histologically proven mass lesions and 129 histologically proven nonmass lesions determined through MR imaging studies. The Breast Imaging Reporting and Data System classification of mass and nonmass enhancement was obtained from radiologic reports.

After extracting 176 features, the researchers ranked them according to kinetic, texture, and morphologic features for mass and nonmass lesions. They found that the best classifier performance was a two-stage cascade classifier, whereby the researchers used mass versus nonmass followed by malignant versus benign classification, as compared with a single, or one-shot, classifier, which was benign versus malignant. “Our proposed two-stage cascade classifier decreases the overall misclassification rate by 12 percent, with 72 of 409 missed diagnoses with cascade versus 82 of 409 missed diagnoses with one-shot classifier,” the authors wrote.

By separating the optimizing features, the researchers found that breast MR imaging with CAD was improved over using a one-shot classifier. They suggested that this may provide an advantage to women who are at high risk for developing breast cancer.