Parallel visualization of a 3D heart model in an heterogeneous computing environment

Abstract

Heart diseases are occurring more than ever in the recent years. A Virtual Heart Model would provide a better understanding of the heart’s function and enables us to identify the defections in the heart. The large dataset of the model requires a huge amount of processing time and cluster computing provides a faster as well as a cost effective way to visualise it. The research carried out introduces a parallel approach in visualising a three dimensional (3D) Virtual Heart Model simulation dataset. The work includes a brief review on the Immersed Boundary Method (IBM) and how they differ from regular Computational Fluid Dynamics (CFD). The dissimilarities motivate the need to show a graphical representation of the IBM. The IBM is used to compute the 3D heart model simulation. We present a cost effective method using off-the-shelf commodity cluster to visualise the IBM dataset. This work describes the techniques used in converting the simulated datasets to be visualised. The heart dataset consists of the cardiac fiber orientation, pressure, velocity, stress value markers. However, only the fiber orientation data is used in the initial stage.Analysis of the visualisation is done to ensure that there are no errors in the modeled system. The parallel visualisation is done in a heterogeneous cluster computing environment to improve the rendering performance. The cluster acts as the server and processes the geometry to be rendered by the client. The parallel visualisation uses the sort-last distributed rendering algorithm to process the IBM dataset. An identical visualization pipeline is then created in each of the processor involved. Each pipeline creates geometry of different partition of the entire dataset. The processed geometry is then collected back at the master node which has to be sent to a client. The rendering performance is compared with other point cloud dataset. The end result reveals that the method proposed is scalable even when the datasets become large. The findings pave way for the larger heart dataset that may consist of the pressure and other value markers to be rendered using the same methodology. Consequently, the visualisation assists in the process of identifying errors in the simulated dataset. This research will also assist in low cost building of visualising biological modeling or Geographical Information System (GIS) dataset in the near future.