Improved matrix inversion in image plane parallel MRI.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Image Reconstruction

Background:

• Current parallel MRI methods like SENSE and CAIPIRINHA rely on specific k-space sampling patterns.
• The choice of k-space sampling significantly impacts image reconstruction accuracy and noise levels.
• Existing methods may not fully leverage object support and coil sensitivity information.

Purpose of the Study:

• To introduce Generalized Unaliasing Incorporating Support constraint and sensitivity Encoding (GUISE), a flexible 3D parallel MRI reconstruction method.
• To propose an efficient, adaptive sampling pattern design for GUISE that incorporates prior knowledge.
• To evaluate the performance of the proposed sampling strategy against conventional methods.

Main Methods:

• Development of the GUISE framework for direct image recovery from arbitrary Cartesian k-space data.
• Design of an adaptive sampling pattern method utilizing object support and coil sensitivity profiles.
• Comparative analysis using simulations and experimental data against 2D SENSE and 2D CAIPIRINHA.

Main Results:

• The proposed adaptive sampling pattern design significantly reduces noise in MRI reconstructions.
• GUISE with the new sampling strategy better exploits coil sensitivity variations and object support constraints.
• Reconstructions excluding non-object regions potentially allow acceleration factors exceeding the number of receiver coils.

Conclusions:

• GUISE offers a versatile approach to 3D parallel MRI reconstruction.
• The adaptive sampling pattern design enhances reconstruction quality and efficiency.
• This method holds promise for accelerated and improved clinical MRI acquisition.