GPGPU Acceleration Algorithm for Medical Image Reconstruction

Medical imaging techniques such as X-ray, Ultrasound, CT and MRI scan are widely used for diagnosis. The 2D medical images from these scans are difficult to interpret because they can only show cross section views of a human body. Interpreting these images requires experts or trained professionals. Reconstructing 2D images into 3D models can help with the interpretation process. However, such model reconstruction is normally time-consuming and costly. It requires high performance computation, such as grid or parallel computing. This research, thus, proposes a high performance 3D reconstruction method using the General-Purpose computation on Graphics Processing Units (GPGPU). The GPGPU has a high computational performance. Parallel computing method on GPU can thus regenerate a model for real time 3D visualization. In other words, the GPU computational speed sufficiently improves the visualization effectiveness of both images and models to the point where a real-time navigation of the data set is possible. In our work, the 3D reconstruction process reconstructs a set of 2D cross-section images and stacks them to generate a volume data, and then transform them into a 3D model. The generated models are then displayed on the user interface developed with OpenGL. Finally, the performance of the GPU acceleration is presented in this paper.