Ultrasound Imaging Simulation Platform Based On Gpu

Ultrasound imaging is one of the three modern medical imaging technologies. It is safe, non-invasive, and good in real-time imaging. Ultrasound imaging is widely used in clinical application. It is the combination of fundamental research in medical ultrasound, new piezoelectric materials, computer information technology and other related aspects of the latest achievements.Simulation of ultrasonic imaging system has an important significance. The simulation system can be used in designing imaging system parameters, designing new beamforming algorithm and focus strategies.So the simulation system is very helpful in research and design a prototype imaging system.Designing a simulation platform for imaging system, complete and detailed understanding of the actual imaging system is necessary. In this thesis, the acoustic phenomenons such as: incidence, scattering, attenuation in ultrasonic acoustic imaging system are introduced. And its physical characteristics that will affect the performance of imaging systems are introduced too. The relation between ultrasonic transducer physical properties and emission ultrasound that is very important in imaging system is described. Through analyzing the structure of digital ultrasound imaging systems, simulation platform is divided into three modules: sensor array setup and launch strategies, beamforming, signal processing. Probe array setup and focus strategy are achieved by Field2 programming. Field2 is the linear sound field simulation software, the simulation base on linear system theory, calculation of the spatial impulse response is the kernel of the program.The calculation of the emission field, scattering field and the sensor output signal are based on the kernel. By setting the sensor, focus strategy and input signal, the program can calculate the complete set, digital front-end data or A-Line data. GPU is a digital chip that used for calculation computer graphics. Comparing to CPU its structure has high parallelism,the TESLA GPU architecture has the SPA module supporting general purpose computing, the parallel programming model, can provide nearly 10 times the CPU's floating point computing ability. CUDA programming model is using to design the beamforming program using the GPU computing resources. Signal processing such as Hilbert transform, log compression, image display are programming by using matlab for convenience.4 group experiments about the sensors, transmitting and receiving focus strategy, B-model imaging, CPU and GPU imaging contrast simulation are verified the effect of the physical characteristics of sensors, focusing strategies, such as amplitude apodization in imaging. Beamforming results prove that the GPU and CPU beamforming have the same effect and the imaging simulation platform based on GPU calculation is feasible.