Research On Fast Computation Of Tissue Motion Based On FPGA

Ultrasound elastography technology can quantitatively calculate the motion field of the tissue,and then obtain quantitative and visualized tissue elasticity information of the lesion area.Ultrasound elastography has developed rapidly since it was proposed.At present,ultrasound elastography is widely used in the diagnosis of human soft tissue lesions,which is of great significance for the clinical diagnosis of human soft tissue cancer and other lesions.The calculation of the tissue motion field is the core of the elastic imaging algorithm.Selecting an accurate motion model and a highly parallel algorithm and a highperformance processing platform are of great significance for the rapid calculation of tissue motion.The combination of affine motion model and optical flow method not only solves the problem of inaccurate tissue motion description from the root,but also improves the parallelism of the algorithm to a certain extent,and further improves the speed of tissue motion estimation.However,the biggest problem of the above algorithms is that the calculation on a general-purpose serial processor is still time-consuming and slow.The main reasons are the large amount of ultrasonic echo data per frame and the algorithm involves a large number of matrix operations,resulting that the speed of processing drops rapidly under software serial,finally which limits its application in practical systems.General-purpose serial processors cannot meet the needs of fast calculation of tissue motion.The fine-grained parallel,reconfigurable and flexible features of FPGAs have made FPGA-based embedded application systems widely concerned.This article starts with the theoretical analysis of the affine optical flow method,and uses the advanced HLS highlevel comprehensive language to design and implement a hardware accelerator for fast calculation of tissue motion with affine optical flow.Then from the perspective of software and hardware co-design and based on ZYNQ heterogeneous acceleration platform,this article design and implement the system-level packaging and testing of optical flow IP core,verifying its the portability and versatility.At the end of this article,through comprehensive experiments,it can be seen that when the size of RF data before and after the compression is 1625 * 449,and performing an affine optical flow calculation on the selected ROI block before and after compression,the speed is an order of magnitude faster than the software serial.Through system-level testing,the tissue motion calculation system completed affine flow calculations on all blocks.The speed from raw data to generating tissue displacement and strain images(before interpolation)is 4 times faster than software serial model.