Author: Phillip Cuculich, Geoffrey D. Hugo, Xiwen Li, Michael T. Prusator, Clifford Robinson, Pamela Samson, Xue Wu 👨🔬
Affiliation: Washington University School of Medicine in St Louis, Washington University School of Medicine in St. Louis, University of Utah, WashU Medicine 🌍
Purpose: Stereotactic arrhythmia radiotherapy (STAR) requires compensation for both respiratory and cardiac motions of the heart. Respiratory 4DCT scans implicitly include cardiac motion and cycle-to-cycle respiratory variation, which distort the image. We develop a model-based correction procedure to extract cardiac and respiratory motion directly from a respiratory 4DCT.
Methods: A post-processing motion correction procedure corrected respiratory 4DCT using cardiac-gated 4DCTs as cardiac motion models. This procedure contains a pre-registration, a registration, and a post-registration process. In the pre-registration, the respiratory and cardiac 4DCT scan were globally aligned to the heart. Respiratory 4DCT phases were separated into stacks based on acquisition time. The possible motion range was selected for each stack based on the slice location and respiratory motion range and applied to each cardiac model. In the registration, the cardiac 4DCT phases were pre-aligned and registered to the stack of slices separately within the selected motion range to recover multiple candidate heart positions during that time point of acquisition. Post-registration, the most suitable location and cardiac phase were selected from the candidates to ensure smooth spatiotemporal motion. The process was evaluated in simulated respiratory and cardiac 4DCT acquisitions using the XCAT phantom. Misalignment, noise, and breathing irregularity were simulated.
Results: This procedure produced a 3D cardiac model with realistic respiratory+cardiac motions. The dice similarity coefficient between the motion corrected volume and the ground truth is 0.86 for the whole heart motion envelope and 0.88 for the left ventricle motion envelope. The misalignment (rotation<3o) slightly reduced the similarity while 1%noise have little effect.
Conclusion: In this study, we developed a post-processing motion correction procedure for simultaneous respiratory and cardiac motions and generate a 3D respiratory+cardiac motion image of the heart shape and position. The motion correction procedure showed promising results in phantom. It can be potentially used for other 4dct dataset.