Towards a multimodal cardiac motion atlas

Statistical cardiac motion atlases provide a valuable tool for the quantification and evaluation of heart disease. Typically, such atlases are formed and applied using data from the same modality, e.g. magnetic resonance (MR) or 3D ultrasound (US). This paper presents a feasibility study on building a statistical cardiac motion atlas using one modality (MR), and embedding a new patient into this atlas using another modality (US). We aim to simulate a clinical scenario in which a high quality motion atlas is formed using MR data, and a new patient can be related to this atlas using only low cost ultrasound data, which is widely available in clinics. A leave-one-out technique was used to quantify the similarity between MR and US-derived displacement vectors in a reduced dimensional atlas space. Results show that all subjects' embedding errors were statistically significantly lower than the pairwise distances between each of the MR-derived training embedded points. This feasibility study illustrates the potential of combining two sources of information (i.e. MR-US) for comparing subjects' motions in a common space.