3D Reconstruction of Intricate Archean Microbial Structures Using Neutron Computed Tomography and Serial Sectioning

Huerta, N J
huerta@geology.ucdavis.edu
Geology Department, University of California, Davis, CA 95616 United States 

Murphy, M A
megmurphy@geology.ucdavis.edu
Geology Department, University of California, Davis, CA 95616 United States 

Natarajan, V 
vijayn@ucdavis.edu
Institute for Data Analysis and Visualization, University of California, Davis, CA 95616 United States 

Weber, G 
ghweber@ucdavis.edu
Institute for Data Analysis and Visualization, University of California, Davis, CA 95616 United States 

Hamann, B 
hamann@cs.ucdavis.edu
Institute for Data Analysis and Visualization, University of California, Davis, CA 95616 United States 

Sumner, D Y
sumner@geology.ucdavis.edu
Geology Department, University of California, Davis, CA 95616 United States

Three-dimensional visualization of intricate microbial structures in rocks is essential to understand the growth of ancient microbial communities. We have imaged and reconstructed the three-dimensional morphology of 2.5-2.6 billion year old intricate microbialites preserved in carbonate using both serial sectioning and neutron computed tomography (NCT). Reconstruction techniques vary with data type and sample preservation. NCT is a non-destructive technique for imaging organic-containing samples with sufficiently high hydrogen concentrations. The resolution of reconstruction is finer than 500 microns. We reconstructed microbialites preserved as organic inclusions in calcite using NCT. Reconstructions are interpreted using volume rendering, segmentation, and an interactive Matlab/visualization environment. Visualizations demonstrate the intricacy of the structures. Noise currently limits automatic growth surface extraction, but growth of structures can be qualitatively evaluated. One of the largest obstacles to date is efficient manipulation of large data sets. Our current visualization approach always renders the supplied data set at full resolution, which requires down-sampling of datasets larger than 256 pixels3 (acquired volume data consists of up to 2048 pixels3) to isolate regions of interest and extract important features. We are exploring the use of multi-resolution techniques that store a dataset at different levels of detail and chose an appropriate resolution during user-interaction. Such an approach will allow us to visualize raw data at full resolution. Serial sectioning and scanning successive horizons provides reconstructions of samples lacking sufficient hydrogen for NCT. This technique destroys the sample and has a lower resolution than NCT. However, intricate networks of microbial laminae surrounded by cement-filled voids can be characterized using this technique. After microbial surfaces are manually interpreted on slices, the images lack noise, allowing clean, but less detailed reconstructions. Serial sectioning reconstruction results in high horizontal but low vertical resolution. Therefore, visualization and surface extraction techniques on a selective subset of the data are customized to accurately reconstruct the intricate structures. Results demonstrate that the ancient structures contain vertical, connected planes that have the same scale and spacing as some modern microbial structures.