Back to all papers

Accelerated Cardiac MRI Using Deep Learning Cine Imaging: Validation in Pediatric Patients both with and without Congenital Heart Disease.

August 31, 2026pubmed logopapers

Authors

Kochav J,Ma J,Jambawalikar S,Fung M,Quarterman P,Shah A,Farooqi K,Majeed A,Jhaveri S,Channing A,Hsien S,Janich M,DiLorenzo MP

Affiliations (3)

  • Division of Cardiology, Department of Medicine, Schneeweiss Adult Congenital Heart Center, Columbia University Irving Medical Center, New York, NY, USA. Electronic address: [email protected].
  • GE HealthCare, Milwaukee, Wisconsin, USA.
  • Division of Cardiology, Department of Pediatrics, Columbia University Irving Medical Center, New York, NY, USA.

Abstract

Conventional breath-held 2D bSSFP cine CMR requires multiple breath holds and prolonged acquisition time, posing challenges for children and patients with congenital heart disease (CHD). Deep-learning (DL)-based reconstruction enables undersampled data acquisition with rapid image completion, offering the potential for substantial time savings while preserving diagnostic accuracy. Fifty pediatric patients (median age 16.1 years [14.7, 17.5], 56% CHD) with two-ventricle physiology were prospectively recruited, and underwent conventional breath-held bSSFP cine imaging and DL-accelerated cine imaging at three acceleration levels: with breath-held acquisitions over 3-RR intervals, 1-RR interval, and 1-RR interval with free-breathing. Acquisition time, image quality, and quantitative ventricular measurements were compared across sequences. Median short-axis cine stack acquisition time was reduced by 66%, 83%, and 89% for the 3-RR, 1-RR, and 1-RR free-breathing sequences, respectively, compared with conventional cine (all p<0.001). Image quality remained diagnostic in 96%, 88%, and 88% of cases, respectively with no statistical difference in image quality score between conventional cine and 3-RR acquisition. Quantitative assessment of myocardial thickening via radial strain demonstrated excellent agreement (ICC = 0.92, 95% CI 0.82-0.96, p < 0.001) between conventional cine and 3-RR acquisition, with only slightly higher measurements. on 3-RR sequences (29.7 ± 6.5% vs. 28.5 ± 6.1%, p<0.001). LV and RV volumes differed by <5 and <10mL/m² across all acceleration levels, LV mass by <3g/m², and LVEF/RVEF by ≤2%. Among CHD patients, a 2% overestimation of RVEF was observed. DL Cine enables rapid and diagnostically robust quantification of biventricular size and function in pediatric patients, including those with CHD. Image acceleration with 3-RR acquisitions optimally balances efficiency and image quality, supporting integration of DL cine imaging into routine pediatric and congenital CMR workflows.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.