3D Ultrasound Visualization System Based GPU Programming

Ultrasound, CT and MIR are important imaging methods for modern medicine. In CT or MRI device, since the volume dataset are obtained directly in the 3D Cartesian coordinate, it is comparative easy to render those volume dataset. While the situations in 3D ultrasound device are more difficult, the space structures of the data received from 3D ultrasound probe are mainly depended on the scanning mode of probe. In this way, it is impossible to use a general system to visualize the date from the probe directly. Moreover, it is not allowed to use the accurate but time consumed algorithm in a real-time rendering system.In the traditional 3D imaging system, the data from the probe needs to be re-sampled and reconstructed before it can be transfer to the 3D visualization pipeline. In the re-sample procedure it requires the CPU to process all the data in the dataset and interpolate them in an obviously larger data array. The reconstruction processing consumes a lot of time and produces an obviously larger dataset, which slow down the whole visualizing procedure. In order to solve this problem, this thesis suggested a new GPU based programming procedure to substitute the reconstruction processing by the texture mapping on the vertexes powered by the programmable rendering pipeline of the modern GPU to accelerate the overall system.Meanwhile, this thesis also gave the algorithms and codes in details to realize the algorithm proposed. This thesis could be used as a reference when designing the 3D ultrasound system.