Technical - Emerging CMR Methods - Acquisition & Reconstruction-Discussion

Video Transcription

Video Summary

The session moved into Q&A, focusing on MRI reconstruction, perfusion imaging, 4D flow, and AI in clinical practice. Hassan explained that 28.8x acceleration was chosen to enable real-time phase contrast imaging, which can capture beat-to-beat variation and benefit arrhythmia and pediatric scans, though testing lower acceleration for accuracy loss would be valuable. He also justified removing residual connections because they retained more noise in complex difference images. Garan described estimating pulmonary arterial pressure by measuring vortex presence across 20 cardiac frames and converting vortex duration into pressure using a published equation, noting limited impact from temporal resolution. Tim discussed strong clinical enthusiasm for AI reconstruction, emphasized the need for broad validation across pathologies, and suggested solutions to data-sharing barriers such as anonymized consented datasets and federated learning. The session ended by stressing the role of professional societies in organizing validation and data repositories.

Keywords

MRI reconstruction

perfusion imaging

4D flow

AI in clinical practice

federated learning