| Summary: | 碩士 === 國立清華大學 === 電機工程學系 === 100 === Research on structures and functions in human brains has long been popular. Scientists are trying to find out connections between neural networks in brains and the related diseases such as Alzheimer's or Parkinson's disease. Drosophila, also called fruit fly informally, is widely used in brain research due to some appealing properties such as its similarities in some brain functions like memorizing and learning things, simplicity compared to human brains and easily culturing characteristic. When analyzing brains from different flies, one should first align the significant organs or neurons inside the brains in order to eliminate variations between different flies, making the analysis easier and more robust. Thus, a volumetric registration process is required to match two volume data of brains. When volumetric registration is applied, we need to specify features, those points should be aligned after the procedure, and deform the brain volume into the target one according to these pre-specified features. In practice, large numbers of feature points are necessitated to obtain a precise registration result. However, great number of features involved implying more time is needed for the registration process, especially for 3D volume data. Notice that there is a great amount of features allocated on the boundary of brain slices, that is, the surface of the volume data. We develop a surface registration framework for drosophila brains which can first align the surfaces of brains, thus reducing the number of features required for the following volumetric registration process. Moreover, combined with our surface registration approach, performance of volumetric registration can be greatly improved since traditional volumetric registration process does not take surfaces into account.
|
|---|
