The main purpose of k-space is to maintain the data acquired during a scan prior to processing the image. In the clinical world, k-space is one level in the creation of the image. It is not seen or stored in regular scans. In the research world often you have a choice in keeping the k-pace, for later processing, or changing the parameters in which it is acquired. Creating an MRI image requires the collection of acquired data. Each acquisition within a new transverse magnetization is created and sampled in k-space in a particular order. K-space is a two or three dimensional entity where there exist phase encoding directions along y or z and a frequency encoded direction along x. Changes to the acquisition method are possible and will be discussed later. These changes can affect the scanning and acquisition time, although potentially leading to artifacts. MRI acquires an image by selecting three different gradients, dedicated to spatial encoding, phase encoding, and frequency encoding.
Figure 2 displays k-space sampling trajectories and actual undersampling pattern. As it can be seen, each value represents a portion of the image and selectively picking areas, for sample the middle, you would have a blurred image with lower resolution. As shown in 6 Figure2 undersampling the k-space would not have a significant effect on quality unless certain essential parts of k-space are undersampled. Altering the acquisitions of k-space will result in different images and artifacts,(Lustig,2008). Also there are different trajectories (Figure 3) that k-space can be sampled to decrease scanning time and avoid certain artifacts like aliasing where there will be a misrepresentation of frequency instead of the actual. One of the most important concepts of k-space is the how the center of k-space contributes to contrast while the edges contribute to resolution.