1350206 - Total polar approach is implemented as a software tool for Efficient CMR

Cardiac MRI, as a dynamic imaging, needs more acceleration to compete with modalities that are more popular in practice [1].  Expediting through undersampling is a well-known technique but generates artifacts that are commonly seen in the entire field of view (FOV).  Recently a full polar approach has been suggested for acquisition/reconstruction in MRI that combines radial acquisition with polar Fourier Transform (PFT) [2,3]. In this approach, the undersampling artifacts are pushed towards periphery and have little effects on the central part of the image. This can be particularly advantageous for CMR since the region of interest (ROI) is located in a small area at the central part of FOV. In this study, we provided a software tool to make this approach easily available. The efficiency of this software was tested over two datasets for cine and tagged cardiac images.

Methods: A graphical user interface (GUI) was developed using Tkinter package in Python. It reads and downsamples the raw data in terms of radial spokes. Images are then reconstructed through PFT and can be visualized in both polar and Cartesian coordinates. Radially acquired data for Cine cardiac MRI from Harvard dataset [4] were reconstructed using the developed software with three different sampling rates to compare the image quality in the central part of the image. To examine the adaptability of the developed technique, a series of radially tagged real-time cardiac cine images were also reconstructed.

Results:

Figure 1 shows the outlook of the GUI software. On a PC computer with Intel(R) Corei7-7700HQ CPU @ 2.80GHz, it took 1023 s to reconstruct 400 images from the raw data (for a high # of 196 spokes/image) and save them.
The reconstructed images of a single slice for three different sampling rates (98, 49 and 28 spokes) are depicted in Figure 2. It is evident that the image quality remains intact over the central part of the image if we go 3.5 times faster (28 rather than 98 spokes) but the radius of this "reduced FOV" is getting smaller proportional to the number of spokes. 
Figure 3 shows the radially tagged images that were acquired in real-time using 21 spokes per frame. The heavy blurring in the circumferential direction didn't harm the resolution of the radial tags on the LV wall.



Conclusion:

The provided software tool was able to readily reconstruct CMR images in a reasonable time. Testing over a large set of data (108 cases), showed the efficiency and robustness of PFT for expediting the CMR acquisition. Considering the normal diameter of the heart (almost 1/4th of the entire FOV), we showed that 3.5 to 4 times of acceleration is simply possible through this full polar approach for CMR imaging. This acceleration is independent from the acceleration that can be provided through other techniques such as parallel imaging or compressive sensing. Versatility of the developed technique/software is evident on reconstruction of real-time radially tagged images which was collected in an independent study.