Rapid MRI Acquisition By Image Reconstruction Augmented with Inter-voxel Constraints
A novel and effective algorithm to reduce the geometry-factor artifacts in parallel magnetic resonance imaging.
Background
Magnetic resonance imaging (MRI) is one of the most versatile and safest imaging modalities with numerous applications in medicine. One of the major shortcomings of MRI is long scan times, which in turn leads to motion artifacts, low patient throughput, and increased costs. One approach to reducing the scan time is Parallel MRI, whereby arrays of radio-frequency receiver coils collect different portions of the data simultaneously. Unfortunately, the signal under-sampling associated with parallel MRI results in prominent artifacts known as the geometry-factor artifacts. These artifacts can significantly reduce the diagnostic quality of the image. In order to avoid such artifacts, in clinical practice, parallel MRI is limited to acceleration factors of 2 or 3. Currently, there is no effective method to eliminate or reduce the artifacts at higher acceleration factors.
Technology Overview
BCH researchers have invented a novel and effective algorithm to reduce the geometry-factor artifacts in parallel MRI. The proposed algorithm is based on exploiting the coupling between the signal intensities in neighboring image voxels. The inventors use these inter-voxel intensity couplings to derive additional constraints that improve the conditioning of the image reconstruction inverse problem. These constraints are obtained by computing the first-order and/or second-order differentials of the imaging equation. These extra constraints allow for disambiguating the image intensity values and significantly reducing the geometry-factor artifacts. As a result, the invention enables higher acceleration factors in parallel MRI, while achieving diagnostically acceptable image quality. Importantly, the proposed algorithm is entirely compatible with existing algorithmic approaches that are commonly used to improve the image quality, such as smoothness and sparsity regularization.
The inventors implemented the proposed algorithm for 2D and 3D image reconstruction. They evaluated the algorithm on MRI data acquired with three different pulse sequences and different numbers of receiver coils. On data from two volunteer subjects, the proposed algorithm outperformed the state-of-the-art reconstruction methods. For 2D reconstruction, the proposed algorithm with an acceleration factor of 4 produced better images than conventional parallel imaging with an acceleration factor of 2. For 3D reconstruction, diagnostically acceptable images were produced at even higher acceleration factors.
Applications
The algorithm can be used for image reconstruction in parallel MRI. It is applicable to various pulse sequences in 2D, 3D, and 4D MRI.
Benefits:
- Enables highly accelerated parallel MRI acquisition, thereby increasing patient throughput and reducing the costs.
- Improves the diagnostic quality of the image by reducing the artifacts and increasing the signal to noise ratio.
- Works for low and high acceleration factors
- Does not require any additional hardware components.
Publications:
Yazdanpanah, A.P., Afacan, O. and Warfield, S.K., Reconstruction Augmentation by Constraining with Intensity Gradients (RACING). ISMRM-2019.
