Research Of Near Real-Time 3D Ultrasound Elastography Based On Mechanical Scanning

Ultrasound elastography is a new method of ultrasound imaging which images tissue by getting its elasticity information, remedying shortages of X-ray, ultrasonography, magnetic resonance imaging and computed tomography that cannot provide direct information on the hard or soft tissues. Ultrasound elastography is non-invasive, cheap and easy to conduct. Due to these advantages, it has become a hot research topic in the field of medical ultrasound imaging.Three-dimensional(3D) ultrasound elastography currently can be divided into two main categories according to the adopted ultrasound probe, namely the elastograhpy with 3D probe and elastograhpy with linear array probe. When using linear array probe, a scanning method needs to be performed to acquire radio frequency(RF) signal frame sequences. The scan and acquisition procedure can be carried out by mechanical scanning or freehand scanning with spatial positioning sensor. Due to freehand operation, probe velocity as well as the pressure is not uniform during freehand scanning. Mechanical scanning won’t have this problem because probe movement and the pressure it applied to tissue surface can be precisely controlled. Data collected through mechanical scanning is rather regular, which leads to the high signal to noise ratio(SNR) strain images. When computing tissue displacements, there are algorithms that only consider axial displacement components, and also algorithms that consider the full displacement components in three directions. The time complexity for computing 3D elastic images is relatively high, usually this procedure is carried out offline. With the rapid development of GPU computing techniques in recent years, more and more high computation tasks are processed on GPU. GPU has much better parallel computing performance than CPU, which makes it quit suitable for realizing real-time 3D ultrasound elastography.This paper firstly analyzes the principles of ultrasound elastography, introduces a variety of data collection methods, 3D image reconstruction method, 3D image rendering methods and several common displacement estimation methods with its parallel realization on GPU. Then a near real-time 3D ultrasound elasticity imaging system based on mechanical scanning is designed and implemented. This system consists of a 3D translating device, a medical ultrasound machine and a high-performance server. The 3D translating device controls the probe to scan along a predefined trajectory, the ultrasound machine collects RF signals and sends them to the server wirelessly, the server receives RF signals and does the calculation on GPU. Experiments have been conducted to test the system performance. Ultrasound phantom experiment results show that the system can generate clear and accurate 3D strain images, human forearm experiment results demonstrate the applicability of clinical usage and GPU acceleration experiment results further validate the fact that this system can achieve near real-time 3D imaging.