Segmentation of mitochondria based on SEM images

Recently, due to the rapid development of scanning electron microscope (SEM) with its very high resolution in three dimensions, we can look more closely into those fine structures related to brain function. However, high resolution will inevitably produce huge amounts of data. Considerable attention has been paid for the automated data analysis, which is massively lagging behind data acquisition. In this paper, we focus on the segmentation of mitochondria for its important role in neural function and its homogeneous shape as preferably corresponding landmarks for image registration. Different from previous methods, we present a coarse-to-fine strategy which utilizes the adaptive boosting (Adaboost) algorithm for detection, multi-layer information fusion for the connection across adjacent sections, and sparse region growing for fine segmentation. Finally we show the 3D visualization in ImageJ. Experimental results on automated tape-collecting ultramicrotome scanning electron microscopy (ATUM-SEM) images demonstrate the effectiveness of our proposed method.