Dynamic contrast enhanced (DCE) MRI is used for breast cancer screening examinations for women of high risk for developing breast cancer. Motion caused e.g. by muscle relaxation or coughing during image acquisition can reduce the interpretability of breast MRI. For scans with strong motion, it might be necessary to repeat the scan, but in typical screening workflow this is only detected by the radiologist when the woman has already left the clinic. As a consequence, the woman might need to be recalled for a repeated...

Data set The breast MRI data set was provided by Radboud University Nijmegen Medical Centre (RUNMC) and extracted randomly from original breast cancer screening data. In total, 491 screening exams acquired with the current screening MR protocol of RUNMC [1] from 449 patients were supplied. An expert radiologist manually annotated the clinical image data concerning motion artefacts focusing on the high resolution images of time point t0 (pre-contrast) and t1 (first post-contrast), as these were assumed to be representative for the whole data set. The...

To evaluate the performance of the motion classifier, 10-fold cross validation at patient level was applied to objectively report prediction accuracy. Test results presented as receiver operating characteristic (ROC) curve and statistical measures are depicted in figure 5. By selecting 0.5 as the classification threshold, it was shown that the classifier obtained a high specificity of 0.965, which means 391 out of 405 cases were classified correctly as showing no-motion. A sensitivity of 0.535 (46 out of 86 moderate and severe cases were detected correctly)...

In this work, an automated approach for motion detection in DCE breast MR images was presented. Based on the manually annotated data set, the algorithm showed to extract meaningful features that allow a robust automatic classification. Such a system can be used to trigger an alert after acquisition to warn the technician that the study quality is too low. This would allow re-scanning the women shortly after the first scan, thus preventing unnecessary call-backs. In the future, more annotations will be acquired and included so...

800x600 Normal 0 21 false false false DE X-NONE X-NONE [1] Platel et al. Automated Evaluation of an Ultrafast MR Imaging Protocol for the Characterization of Breast Lesions. Annual Meeting of the Radiological Society of North America 2012. [2] Rühaak et al. Highly Accurate Fast Lung CT Registration. SPIE Medical Imaging, vol. 1, 2013. [3] Breiman. Random Forests. Machine Learning 2001.