Pediatric coronary MR angiography with a two-minute scan using de-aliasing regularization based compressed sensing.
Authors
Affiliations (6)
Affiliations (6)
- National Engineering Research Center of Advanced Magnetic Resonance Technologies for Diagnosis and Therapy, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China.
- Department of Radiology, Children's Hospital, Fudan University, Shanghai, China; National Engineering Research Center of Advanced Magnetic Resonance Technologies for Diagnosis and Therapy, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China.
- Department of Radiology, Children's Hospital, Fudan University, Shanghai, China.
- United Imaging Healthcare, Shanghai, China.
- Department of Radiology, Children's Hospital, Fudan University, Shanghai, China. Electronic address: [email protected].
- National Engineering Research Center of Advanced Magnetic Resonance Technologies for Diagnosis and Therapy, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China; Department of Radiology, Children's Hospital, Fudan University, Shanghai, China. Electronic address: [email protected].
Abstract
Long acquisition time limits the clinical utility of coronary magnetic resonance angiography (CMRA) in pediatric populations. While deep learning-based reconstruction methods such as De-Aliasing Regularization-based Compressed Sensing (DARCS) hold promise for accelerating CMRA, its clinical feasibility in pediatric populations remains unexplored. This study aims to reduce scan time and evaluate the image quality and diagnostic performance of DARCS-accelerated CMRA in pediatric coronary imaging, with a focus on coronary artery aneurysms (CAAs) detection. A two-phase study including retrospective technique development and prospective clinical validation was performed. CMRA was performed using a 3.0 T scanner with a three-dimensional diaphragm-navigated, T2 prepared gradient echo sequence. In the Phase I, pediatric CMRA k-space data were retrospectively undersampled to train and test DARCS reconstruction, with comparison to SENSE, patch-based reconstruction (PROST), and hybrid deep-learning iterative reconstruction (hybrid DL-IR). In Phase II, patients prospectively underwent both conventional 3× accelerated CMRA and 8× DARCS-CMRA. Images were assessed using quantitative image metrics (peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM)), coronary artery assessments (vessel lengths, sharpness, and visual scores). Diagnostic performance for CAA detection was evaluated at both patient and vessel levels. A total of 123 pediatric patients were included for final analysis, including 96 for the retrospective phase and 27 for the prospective phase. DARCS outperformed the second highest-performance reconstruction method in PSNR (31.74 ± 2.17 vs. 30.69 ± 2.12, P < 0.001 at 8× acceleration), improved vessel length (LAD: 75.86 ± 20.17 mm vs. 72.23 ± 20.80 mm, P < 0.001; RCA: 84.94 ± 20.36 mm vs. 80.12 ± 20.54 mm, P < 0.001), and improved subjective scoring (LAD: 3.22 ± 0.83 vs. 3.11 ± 0.89, P = 0.102 > 0.05; RCA: 3.53 ± 0.81 vs. 3.42 ± 0.81, P = 0.046 < 0.05). In the prospective phase, DARCS-CMRA achieved a 100% sensitivity and specificity in detection of CAA at both patient and vessel levels, with conventional CMRA as the reference, despite a significantly shorter scan time (92.4 ± 19.1 s vs. 208.8 ± 52.0 s). DARCS offers improved reconstruction quality for accelerated CMRA compared to conventional methods, enabling preservation of CAA diagnostic accuracy despite a two-minute scan.