Bocugoz E.Erdil E.Argunsah A.O.Unay D.Cetin M.2023-06-162023-06-1620179.78E+12https://doi.org/10.1109/SIU.2017.7960482https://hdl.handle.net/20.500.14365/360325th Signal Processing and Communications Applications Conference, SIU 2017 -- 15 May 2017 through 18 May 2017 -- 128703Analyzing morphological and structural changes of dendritic spines in 2-photon microscopy images in time is important for neuroscience researchers. Correct segmentation of dendritic spines is an important step of developing robust and reliable automatic tools for such analysis. In this paper, we propose an approach for segmentation of 3D dendritic spines using nonparametric shape priors. The proposed method learns the prior distribution of shapes through Parzen density estimation on the training set of shapes. Then, the posterior distribution of shapes is obtained by combining the learned prior distribution with a data term in a Bayesian framework. Finally, the segmentation result that maximizes the posterior is found using active contours. Experimental results demonstrate that using nonparametric shape priors leads to better 3D dendritic spine segmentation results. © 2017 IEEE.trinfo:eu-repo/semantics/openAccess3D dendritic spine segmentationlevel setsnonparametric shape priorsParzen density estimatorImage segmentationBayesian frameworksDendritic spineLevel SetParzen density estimationParzen density estimatorPosterior distributionsSegmentation resultsShape priorsSignal processingConference Object10.1109/SIU.2017.79604822-s2.0-85026309893