Automated atlas-based segmentation of NISSL-stained mouse brain sections using supervised learning

摘要: The problem of segmentation mouse brain images into anatomical structures is an important stage practically every analytical procedure for these images. present study suggests a new approach to automated in the NISSL-stained histological sections brain. algorithm based on method supervised learning using existing labeling corresponding from specialized atlas. A section be segmented preliminarily associated with atlas displaying maximum similarity. image this then preprocessed order enhance its quality and make it as close possible. An efficient luminance equalization, extension well-known Retinex proposed. random forest trained pixel feature vectors constructed class labels extracted labeling. classifier applied classify pixels experimental structures. combination features superpixels location priors suggested. Accuracy obtained result increased by Markov field. Procedures equalization subsequent have been tested real sections.