Detection and Prediction of Neural Structures in Fluorescence Images Using Correlative Light/Electron Microscopy
Software development to analyze biological data becomes more and more important in recent years and is in focus of this project. Fluorescence imaging enables neuroscientists to track activity-dependent changes over time. Therefore, the detection of spines is required. This is a very tedious and error-prone task. Furthermore, the bias introduced by a manual expert analysis can influence results. Therefore, an automated spine analysis will improve the process of spine detection with respect to time and reproducibility.
Within this project some classical approaches were studied, compared and a novel approach implemented. With our technology first automated analysis could be performed. Spine detection without any a-priori knowledge is very limited as could be experienced during this project. Therefore, a new concept based on a correlative dataset was proposed. The correlative dataset is generated by 2-photon imaging and serial block-face scanning electron microscopy. In serial block-face scanning electron microscopy the spines are resolved in all details and a high a-priori knowledge can be gained.
Within this project the advantages of correlative datasets could be demonstrated and a new concept for image analysis of light microscopy data be motivated.
Keywords: Spine Detection, Fluorescence Image Analysis, Correlative Microsopy, Statistical Modelsback