Related Experiment Video
Updated: Aug 5, 2026
07:17
Leveraging Virtual Reality for Immersive Segmentation and Analysis of Cryo-Electron Tomography Data
Published on: January 24, 2025
High Resolution Isotropic 'Pseudo' 3D Cine imaging with Automated Segmentation using Concatenated 2D Real-time
Mark Wrobel1, Tina Yao1, Ruaraidh Campbell1
1UCL Centre for Translational Cardiovascular Imaging, University College London, 20c Guilford St, London, UK, WC1N 1DZ.
Summary
Deep learning models transform 2D real-time cardiac MRI into 3D datasets for faster pediatric heart imaging. This novel approach offers diagnostic quality pseudo-3D cardiac magnetic resonance (CMR) data in under a minute.
Area of Science:
- Cardiovascular imaging
- Medical artificial intelligence
- Pediatric cardiology
Background:
- Conventional cardiovascular magnetic resonance (CMR) uses 2D breath-hold and static 3D imaging.
- Pediatric and congenital heart disease assessment requires efficient imaging techniques.
- Current methods can be limited by breath-holding requirements and acquisition time.
Purpose of the Study:
- To develop a Deep Learning (DL) method for creating isotropic, fully segmented 3D-cine datasets from 2D free-breathing real-time cine images.
- To assess the feasibility of generating 'pseudo' 3D-cine data for pediatric cardiovascular imaging.
- To reduce acquisition and processing time in cardiac magnetic resonance imaging.
Main Methods:
- Trained four DL models for interslice signal/respiratory correction, super-resolution, and anatomical segmentation (atria, ventricles, great vessels).
- Validated the method by converting prospectively acquired sagittal 2D real-time cine stacks into segmented pseudo 3D-cine data in 20 patients.
- Compared quantitative metrics (volumes, diameters) and image quality against reference-standard breath-hold cine and 3D whole-heart imaging.
Main Results:
- Successfully transformed all real-time data into pseudo 3D-cines with offline reconstruction and post-processing under 1 minute.
- Demonstrated no significant bias in left/right ventricular volumes (EDV, ESV) with reasonable agreement and correlation.
- Showed reasonable agreement for vessel diameters, with minor overestimation in main and right pulmonary arteries; DL pseudo-3D-cine data achieved adequate diagnostic quality.
Conclusions:
- Developed a DL-based technique to generate pseudo 3D-cine cardiac magnetic resonance data from 2D real-time acquisitions.
- The method offers rapid acquisition and reconstruction times (<1 min) with fully segmented outputs.
- Open-source trained models facilitate sharing, potentially accelerating clinical CMR practice through improved efficiency and diagnostic quality.
Keywords:
3D-cineCongenital heart diseaseDeep-LearningFree-breathingNon-contrastPediatric cardiologyReal-time
