An Embedded Ultrasonic Data Acquisition And Processing System

Ultrasound medical imaging is widely used in clinical diagnosis. It has theadvantages of low-cost, real-time, small damage. With the development of science,new technology such as MRI, CT can cooperate with Ultrasound medical image byusing image-guided, image-registration, and image fusion for better diagnostic quality.Higher demands on the ultrasound image quality by flitting out noise, restores theoriginal signal, improving the SNR (signal to noise ratio) became an importantprocess in ultrasound imaging. With OMAP-L138as the core embedded processor,this article propose and develop a portable ultrasonic imaging data acquisitionhardware system and algorithm for image denoising. Major research and innovationare as follows:Embedded system design for ultrasonic data acquisition and image processing.In this part, embedded system is designed based on completely analyze of theultrasonic data acquisition system’s function and requirements. A solution that TI’sOMAP architecture processor act as the core control and data processing part isproposed. OMAP-L138processor integrate an ARM9core and a C674x DSP core.The ARM9core with the frequency of456MHz can easy run the embedded operatingsystem for system control. The DSP core focus on algorithm of image processing.This architecture can greatly increase the efficiency of the system. After completion ofthe overall program design, functional modules, including system memory(NANDFlash and DRAM), amplification circuit, VGA circuit, power circuit, as well as theentire PCB layout is appropriately designed for ultrasonic data acquisition. In the end,software running on the hardware is developed under the CCS(Code Composer Studio)environment, and the algorithm running on DSP is transplanted to the embedded hardware by Matlab DSPbuilder tool.Ultrasound image denoising algorithm.Speckle noise is the main noise which seriously affects the quality of ultrasoundimages. Speckle noise filtering is an important step for medical ultrasound imageprocessing. In this paper, a new curvelet threshold method based on the speckle noiseprobability distribution is proposed. Carotid ultrasound images with Gaussian noiseand speckle noise at different level was tested to compare the performance of curveletdenoising method with others, such as median filtering, wiener filtering and db5wavelet. Furthermore, threshold function proposed here is also compared withhard-threshold, soft-threshold, semisoft-threshold and asymptotic-threshold methods.The results show that curvelet denoising method based on noise probabilitydistribution have stable performance at different noise level. The method always hasbetter mean structural similarity index map (MSSIM) and increase the peak signal tonoise ratio (PSNR) about1~2dB, validating the effectiveness of the algorithmproposed in the part.