Purpose: High temporal resolution volumetric thoracic MRI is needed to accurately describe the lung and lung tumor motion. The main aim of the present study is to substantially increase 3D acquisition speed of lung MRI with k‐space under‐sampling along the phase encoding directions and subsequent image reconstruction using Compressed Sensing (CS). Method and Materials: CS was simulated using fully sampled thoracic breath‐ holding MRI of a lung cancer patient. The imaging resolution is 1.8×1.8×2.6 mm. Biased random under‐sampling was imposed in the k‐space by selecting more points near the center, and fewer points near the bottom left and bottom right corners. An unbiased random under‐sampling scheme was performed for comparison. The data was reconstructed using the Split Bregman method for L1 regularized problems, which solve the unconstrained optimization problem iteratively by spliting the L1 and L2 components. The method removes the artifacts due to the non‐uniform under‐sampling by minimizing the total variation (TV) while maintaining fidelity with the sampled measurements. Results: We were able to reconstruct the data sub‐sampled with different under‐sampling factors (2×2 − 4×4 folds). Images reconstructed with very low sampling ratio of 13.05% were acceptable. The biased under‐sampling strategy required 30% fewer samples than the unbiased under‐sampling technique for equivalent results. The overall quality and resolution of the reconstructed image was comparable to the fully sampled dataset in the conspicuity of fine lung features and organ shapes, which are degraded by noise in images reconstructed from same dataset without using the CS method. Conclusions: We have successfully demonstrated the application of CS to high‐resolution 3D MRI of the lung, providing an acceleration factor up to 4x4. While higher acceleration factor is needed for real‐time full thoracic imaging, the current method would enable real time imaging of a subvolume of the lung for regional motion study.
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging