Abstract
In the research of life science, a fruit fly, Drosophila melanogaster, with the abilities of learning and memory is chosen for research to facilitate the understanding of structures and functions of the brain neural network. A standard brain model of a fruit fly is very helpful for related researches. A fruit fly’s brain is composed of many neuropils. The standard models of different neuropils have been built. To build the complete standard brain model, the relative position and orientation of these standard models of neuropils should be established. This requires a model based registration algorithm. The model is composed of surfaces which are made of many triangles. Therefore the registration algorithm is surface based. A distance map will be helpful for the surface-based registration. A fast algorithm will be revised in this study. Then a two-stage algorithm includes global and local registration will be proposed. The global registration is based on affine transform, which is used to eliminate the global variations between these models. And a local registration named grid registration will be proposed to provide fine, detailed registration between these models. After the surface-based registration, the complete standard brain model will be established. It will be used to register the original neural networks of individual fruit flies’ brains which are composed of a set of volume data. This is a registration between surface model and volume data. An edge detection algorithm will be utilized to extract the brain surface information from the set of volume data. The edge detection results will be used to register with standard brain surface model. Simulation results will show that the algorithm is really useful and reasonable for the registration of fruit flies’ brains.