Reference-Free 2D-to-3D Multiple Slice Transport Average Registration (MSTAR) for Late Gadolinium Enhancement Cardiac MRI

We present MSTAR, a reference-free statistical method for reformatting 2D image slices into a coherent 3D volumetric dataset using sliding window technique and transport averaging based registration. This approach handles motion and anatomical changes, enabling enhanced visualization quality and coherence in free-breathing cardiac MRI. In a study of 74 patients with intracardiac devices, MSTAR generated high-quality 3D volumes without a pre-existing target, achieving 81.3% sensitivity and 92.2% specificity compared to standard 2D protocol. The 3D method enabled LGE findings confirmation in multiple views, eliminating indeterminate segments present in 2D protocol. This technique bridges the gap between 2D and 3D LGE imaging, improving diagnostic confidence in complex cases and benefiting patients unable to hold their breath for segmented scans.