Free-Breathing & Ungated Cardiac MRI Using Iterative SToRM (i-SToRM)

Area of Science:

• Medical Imaging
• Image Reconstruction
• Magnetic Resonance Imaging

Background:

• Dynamic MRI data acquisition is often limited by scan time, necessitating undersampling.
• Existing methods like SmooThness Regularization on Manifolds (SToRM) utilize global image manifold structures.
• Navigator-based manifold estimation, used in SToRM, is not feasible for all dynamic MRI scenarios.

Purpose of the Study:

• To introduce a novel local manifold regularization approach for dynamic MRI.
• To enable faster MRI scans by reducing data requirements.
• To overcome limitations of navigator-based manifold estimation in specific MRI settings.

Main Methods:

• A local manifold regularization scheme is proposed, focusing on local image patch structures at specific spatial locations.
• The method generalizes the SToRM scheme by adapting to spatially varying manifold properties.
• A reformulated SToRM approach is used, with the regularization term based on robust distances between image sub-patches, and an alternating optimization algorithm.

Main Results:

• The proposed scheme effectively recovers dynamic MRI data from highly undersampled measurements.
• It significantly reduces data demand compared to global manifold approaches.
• Demonstrated utility in in-vivo prospective free-breathing cardiac CINE MRI and simulated phantoms.

Conclusions:

• The local manifold regularization approach offers a significant reduction in MRI scan time.
• This method facilitates recovery from shorter scans by exploiting local image manifold structures.
• The reformulated SToRM approach provides a viable alternative for dynamic MRI reconstruction where navigator-based methods are unsuitable.