The heart is a challenging organ to image with MRI, as it is not only in continuous motion, but it is also located at slightly varying locations and angles in different patients. Cardiac MRI typically involves acquiring multiple localizer views at various angles in order to arrive at planes that are perfectly aligned to the axes of the heart instead of the normal coronal, sagittal and axial planes. This requires a significant amount of time and specifically trained MR technicians to perform these exams. At the recent Society for Cardiovascular Magnetic Resonance (SCMR) conference, Siemens showed how MR software can automatically plan the basic long-axis views of the heart, with further views probably coming in the future.
After rough localization of the heart within the thorax, the software, called AutoAlign Heart, acquires a stack of slices covering the entire heart in the approximate short-axis orientation using a single-shot TrueFISP sequence. A machine learning-based algorithm, trained on 517 patient datasets, then localizes and segments the left ventricle and detects five anatomical landmarks. Based on these landmarks, the long-axis views of the heart are calculated automatically and then acquired in end-diastole using a single-shot TrueFISP sequence.
The software was tested in 27 patients undergoing cardiac MRI for various reasons. Long-axis views were acquired twice, first using automated planning and subsequently using manual planning. A radiologist assessed the two images for visual differences and graded overall agreement. The algorithm successfully found the heart in all cases. One major error in view planning was encountered with both manual and automatic planning. The error with automatic planning was encountered in a patient with congenital heart disease, a situation for which the software was not yet trained. In more than 90% of cases, there was no clinically relevant difference between the manual and automatic planning methods.
More at Aunt Minnie Europe: Cardiac MR software automatically plans long-axis heart views…