Abstract
Purpose: :
The collection of histological sections of the human orbit, created forty years ago by Koornneef, Los and de Haan is the largest and most detailed study of human orbital anatomy to date. The sections have been digitised. Visualizing this large dataset in 3D, accessible via internet, would enhance insight into orbital anatomy worldwide. Two main problems in visualizing this dataset have been identified. First, enabling interactive online reconstruction of the dataset is a challenge because of limited bandwidth and processing power for this massive data. Also, the interpolation to generate oblique sections is difficult in this case, as the slices are 60-100µm thick.
Methods: :
A client-server system is being developed. As the dataset consists of high-resolution sections (10 µm/pixel), large amounts of data have to be processed in order to provide visualizations to clients. Therefore, the system is designed to be scalable, using multiple servers to allow distributed (multiple computers), parallel (multiple processes) rendering.To provide a reliable representation of the data, several techniques have been selected for comparison: Multi-Planar Reconstruction, for viewing arbitrary slices through the data, Direct-Volume Rendering for viewing an entire specimen, or a part thereof, in 3D, and Iso-Surfacing for three dimensional viewing of tissue by generating surfaces.
Results: :
The selected system design allows adding arbitrary numbers of computing nodes to a collection of servers. Techniques are successfully implemented to progressively send resulting imagery to clients using low-bandwidth internet connections. As soon as the first results are available, they are transferred to the client, and refined while the user is waiting. Using these techniques resulted in response times below the interactivity threshold (8 to 10 seconds).Reliable interpolation between reconstructed planes remains difficult (digitization will discard details available using a microscope) but the most appropriate interpolation technique available has been chosen. In addition, providing the user with a measure for reliability of reconstructed planes has proven useful.
Conclusions: :
We have previously shown that visualizing our dataset and providing reliable, reconstructed visualizations is possible. We aim to provide internet access to detailed human anatomy available for researchers in the public domain using common broadband connections.
Keywords: orbit • image processing • microscopy: light/fluorescence/immunohistochemistry