Application of FPGA in Medical 4D Imaging

Medical imaging is one of the most valuable tools that doctors have in detecting and diagnosing disease or abnormality in their patients. From ultrasound, which provides fast 2D images, to computed tomography (CT) and magnetic resonance imaging (MRI), which provide highly accurate 3D images of the human body, both 2D and 3D imaging play an important role in enabling medical professionals to provide better clinical outcomes. However, the advent of 4D imaging is taking medical imaging to the next frontier of moving images. For example, 4D imaging is used in the MRI respiratory analysis cycle. Such advancements in medical imaging are exciting, but not without their challenges. 4D MRI imaging requires extensive pre- and post-processing to reconstruct the image.

An MRI scan consists of two elements: the scan, which is performed during the acquisition of data, followed by reconstruction. During the scan, data samples are captured along a predetermined trajectory. These samples are spatial in nature and in the so-called k-space domain. The acquired samples are converted into an understandable image during the reconstruction phase. MRI therefore faces the competing challenges of producing high-definition imaging, low signal-to-noise ratio, and fast scan times.

The complexity of image reconstruction depends on the sampling trajectory. Simple Cartesian scan directions provide k-space samples that are aligned to a grid, allowing for rapid reconstruction of images using Fast Fourier Transforms. Non-Cartesian scans, such as spiral trajectories, will result in k-space samples aligned in a more complex pattern, which requires advanced image reconstruction algorithms. Currently, it can take several minutes for an image to be available after a scan is completed and requires considerable processing power. This makes it difficult to deploy 4D imaging solutions, hindering wider adoption. For research purposes, servers can be used to demonstrate algorithm performance. However, a deployable solution requires computing power capable of performing RF signal drive, signal capture, and image reconstruction.

Programmable SoCs are a solution that enables widespread adoption of 4D imaging. Xilinx heterogeneous Zynq UltraScale+ MPSoCs help address the various challenges facing 4D imaging by taking advantage of an integrated high-performance processor system with parallel programmable logic.

Thanks to the unique architecture of these devices, the programmable logic can be used to interface with the RF drive waveform and ADC to capture the resulting data from the scan. This extensive interface capability also enables parallel data structures to be implemented within the programmable logic to support multiple parallel high-speed RF generation or signal capture. At the same time, the processing system can be used to generate the user interface, communicate with the medical record system, and more.

Both the RF drive waveform and image reconstruction can be accelerated using programmable logic resources. Because these algorithms are complex, developers can improve productivity by using Xilinx Vitis, a high-level synthesis (HLS) tool that allows engineers to develop algorithms in C/C++ or OpenCL without having to work at the hardware description level. Using Vitis HLS, developers can define algorithms at a higher level and take advantage of the parallel nature of programmable logic, such as unrolling loops and pipelining, to exploit the parallelism that exists in the algorithm. Implementing algorithms in programmable logic can significantly improve performance.

Summary

Medical imaging is a key technology that enables medical professionals to gain a more detailed understanding of the human body and perform advanced treatments, but technologies such as 4D imaging require increased computing performance. Xilinx heterogeneous Zynq UltraScale+ MPSoCs and advanced tool chains are able to support RF drive, signal capture, and image reconstruction, enabling faster scans and widespread adoption of 4D.