Frontal slice from a 3-D confocal laser-scanning microscopy (CLSM) image of an individual bee brain.
Frontal slice from a 3-D confocal laser-scanning microscopy (CLSM) image of an individual bee brain.
This project is a multi-year collaboration (since 2000) with the group of Prof. Randolf Menzel at the Institute of Neurobiology, Freie Universität Berlin (Berlin, Germany). We are interested in applications of various image analysis techniques to confocal laser scanning microscopy (CLSM) images of bee brains. Such images show an entire bee brain in high resolution (approx. 4μm pixel size), following an immuno-histological fluorescent staining procedure.
Frontal slice from a 3-D magnetic-resonance microscopy (MRM) image of an individual bee head capsule with brain.
In our most recent work, we have investigated the spatial deformations that a bee brain undergoes during the specimen preparation process for CLSM imaging. We compared CLSM images from 20 bee brains to an in situ magnetic resonance micrscopy (MRM) image of a brain, acquired by Frank Schaupp and Daniel Haddad in the Experimental Physics Department V at the University of Würzburg (Würzburg, Germany) using an 11.75T magnet. Since the head capsule remains intact during MRM imaging, and because the chemical treatment of the brain is less aggressive, the brain can be expected to remain very close to its in vivo shape.
Mean deformation field between 20 individual bee brain CLSM images and a reference individual MRM image. Image courtesy of Daniel Haddad.
One ultimate goal of our work is to be able to correct the spatial transformations that a brain is subjected to during the preparation process, and unwarp it back into its original in vivo shape. Since the CLSM and MRM images that are currently available to us originate from different individuals, we cannot quantify the actual deformation of a single individual brain. We can, however, analyse the mean deformation of all 20 individual brains that were imaged in CLSM. The mean deformation field shows a clear contraction (about 15%) in the lateral direction. It also shows an increase in brain thickness of approximately the same magnitude.
CLSM image with contours from atlas-based image segmentation.
In earlier work, we applied atlas-based segmentation methods to label compartments in the bee brain automatically. We found that the accuracy of the automatic segmentation (compared to a manual ground truth segmentation) was substantially improved by using more than a single atlas, or by using an average shape atlas.
URL: http://www.stanford.edu/~rohlfing/research/bee/index.html
Last updated October 19 2009 11:45:29.
Torsten Rohlfing, Ph.D., torsten@synapse.sri.com
SRI International, Neuroscience Program
333 Ravenswood Avenue,
Menlo Park, CA 94025-3498, USA