ABSTRACT
Spatiotemporal MRI techniques acquire a time-series of images that encode physiological information of clinical interest, such as organ motion [2], contrast agent uptake [1, 17], signal relaxation [30], water diffusion [18], spectroscopy [6], and neural activity based on blood oxygenation [4, 5] (Figure 17.1). The rich diversity of physiological information that can be acquired using this user-defined time-dimension is a unique feature of MRI. However, the relatively limited imaging speed of conventional MRI imposes limitations on the amount of information that can be acquired per unit time. Hence, conventional methods sacrifice spatial resolution and/or volumetric coverage for an appropriate temporal resolution. In reality, because of extensive spatiotemporal correlations, the information content is much lower than the number of pixels times the number of frames. An immediate consequence is that a conventional MRI method, operating by acquiring fully sampled images at each time point, is rather wasteful. Ideally, one would just want to acquire the ‘innovation’ from one frame to the next.
