3D Reconstruction of Neural Circuits from Serial EM Images
b'A basic requirement for reconstructing and understanding complete circuit diagrams of\\nneuronal processing units is the availability of electron microscopic 3D data sets of large\\nensembles of neurons. A recently developed technique, "Serial Block Face Scanning Electron\\nMicroscopy" (SBFSEM, Denk and Horstmann 2004) allows automatic sectioning and\\nimaging of biological tissue inside the vacuum chamber of a scanning electron microscope.\\nImage stacks generated with this technology have a resolution sucient to distinguish different\\ncellular compartments, including synaptic structures. Such an image stack contains\\nthousands of images and is recorded with a voxel size of 23 nm in the x- and y-directions\\nand 30 nm in the z-direction. Consequently a tissue block of 1 mm3 produces 63 terabytes\\nof data.\\nTherefore new concepts for managing large data sets and automated image processing\\nare required. I developed an image segmentation and 3D reconstruction software, which\\nallows precise contour tracing of cell membranes and simultaneously displays the resulting\\n3D structure. The software contains two stand-alone packages: Neuron2D and Neuron3D,\\nboth oering an easy-to-operate graphical user interface (GUI).\\nThe software package Neuron2D provides the following image processing functions:\\n\\u2022 Image Registration: Combination of multiple SBFSEM image tiles.\\n\\u2022 Image Preprocessing: Filtering of image stacks. Implemented are Gaussian and\\nNon-Linear-Diusion lters in 2D and 3D. This step enhances the contrast between\\ncontour lines and image background, leading to a higher signal-to-noise ratio, thus\\nfurther improving detection of membrane borders.\\n\\u2022 Image Segmentation: The implemented algorithms extract contour lines from the\\npreceding image and automatically trace the contour lines in the following images\\n(z-direction), taking into account the previous image segmentation. They also permit\\nimage segmentation starting at any position in the image stack. In addition, manual\\ninteraction is possible.\\nTo visualize 3D structures of neuronal circuits the additional software Neuron3D was\\ndeveloped. The program relies on the contour line information provided by Neuron2D to\\nimplement a surface reconstruction algorithm based on dynamic time warping. Additional\\nrendering techniques, such as shading and texture mapping, are provided.\\nThe detailed anatomical reconstruction provides a framework for computational models\\nof neuronal circuits. For example in \\nies, where moving retinal images lead to appropriate\\ncourse control signals, the circuit reconstruction of motion-sensitive neurons can help to\\nfurther understand the neural processing of visual motion in \\nies.'