VISTA MRI Sampling Using Gradient Descent for Temporal Resolution

The patent applies dynamics by making the sampling pattern adaptive rather than static. The VISTA method dynamically adjusts the sampling distribution across different time frames while maintaining a consistent number of samples per frame, allowing the system to adapt to temporal variations while preserving temporal resolution consistency. This is achieved through an iterative optimization process that enforces temporal constraints during sampling pattern generation.

The patent changes the sampling density parameter across different regions of k-space while maintaining a fixed number of samples per time frame. By varying the spatial distribution parameters rather than the temporal sampling rate, the method achieves both uniform coverage and consistent temporal resolution. The variable density weighting function modifies sample probability density without altering the temporal sampling structure.

The patent transforms purely random sampling into pseudo-random sampling by introducing a variable density weighting function that modifies the probability distribution of sample locations. This parameter change ensures that while samples remain randomly distributed (maintaining incoherence), their density is regulated to prevent excessive gaps or clustering. The weighting function acts as a constraint that shapes the random distribution without eliminating its incoherent properties.

The patent implements feedback through an iterative optimization process that evaluates the sampling pattern and adjusts it to meet both incoherence and consistency requirements. The method calculates metrics such as sample density distribution and incoherence measures, then refines the sampling pattern accordingly. This feedback loop ensures that the final sampling pattern achieves the desired balance between incoherence and result consistency.

The patent makes the sampling method dynamic and adaptable by allowing different constraint types to be incorporated through the optimization framework. The VISTA method can enforce various domain-specific constraints such as fixed temporal resolution, fully-sampled time-averaged k-space, and controlled eddy currents by modifying the objective function and constraints in the iterative optimization process. This dynamic adaptability maintains sampling uniformity while accommodating diverse application requirements.

The patent segments the sampling design into multiple independent constraint components that can be individually enforced. The method separates spatial uniformity requirements from temporal resolution requirements and other domain-specific constraints, allowing each to be addressed independently through the optimization process. This segmentation enables flexible incorporation of different constraints without compromising overall sampling uniformity.

The patent changes the sampling density parameter across k-space regions to optimize image quality at high acceleration rates. By using variable density sampling that concentrates samples in central k-space regions while reducing density in peripheral regions, the method maintains image quality even with reduced total samples. This parameter change allows higher acceleration rates without proportional degradation of image quality.

The patent applies local quality by differentiating sampling density across different regions of k-space. The central region receives higher sampling density to preserve image quality and reduce artifacts, while peripheral regions receive lower density. This localized quality adjustment allows the system to achieve high acceleration rates while maintaining critical image quality metrics through selective sampling distribution.