Reading Volume Datasets from Storage – Using Segmentation Metadata, for an Enhanced User Experience
Branislav Madoš, Norbert Ádám · Acta Polytechnica Hungarica · 2021
This paper deals with the issues of volume dataset representation as an important part of data storage and processing in many fields including science, research and development, medicine or industry.Due to the significant amount of data included in volume datasets, operations performed on them are often, time-and space-consuming.One of those operationsloading data from secondary storage into the operating memory of computer or memory of graphics cardcan be time-consuming and lead to a bad user experience and significantly delay the subsequent processes.Therefore, the main contribution hereof is the design and introduction of an algorithm to generate volume dataset segmentation metadata.It allows (with a small data overhead, as a trade-off) to prepare metadata about splitting the particular volume dataset into segments with different priority levels.Subsequently, it is possible to reorganize the volume dataset according to the priority of the data segments, in descending order.The algorithm proposed herein allows to start the visualization of the volume dataset in its final quality (resembling visualization of the complete volume dataset, although only a part of the data was loaded from the secondary storage), within a fraction of the total load time of the volume dataset.The remaining data are continually read in the background during data visualization, without affecting volume data visualization quality.The first section herein, contains an introduction to the proposed algorithms.Results of tests, performed with different parameter setups on non-invasive medical imaging volume datasets, obtained by computed tomography and magnetic resonance imaging, are included in the second part of the paper.Conclusions, drawn from test results, are summarized in the last part of the paper.