Tez

Permanent URI for this collectionhttp://acikerisim.bau.edu.tr:4000/handle/123456789/160

Browse

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Dendritic spine segmentation using active contour with shape prior
    (Bahçeşehir Üniversitesi Fen Bilimleri Enstitüsü, 2012-06) Akarsu, Cemal; Ünay, Devrim
    Dendritic spine analysis is a popular topic in Neuroscience, since the morphological and statistical (e.g. quantity and density) properties of dendritic spines play a role in the learning process, and can be used for understanding the reasons of mental disorders like Alzheimer’s. Advancements in the related imaging technology has lead to vast amount of dendritic spine images that need to be analyzed and interpreted by the experts. But manual analysis of these images is a difficult and time-consuming task. Therefore, automated tools are required, which should detect, segment, and quantify statistical properties of dendritic spines, such as length and volume. To this end, a collaborative effort on creating a software tool for automated analysis of dendritic spines is being carried out at Bahçeşehir University. This thesis is focused on the automated segmentation part of this collaborative effort. For accurate segmentation of dendritic spines, application of the Active Contour with Shape Prior method is proposed. Performance of the proposed method is evaluated on an expert-annotated dataset, and compared with the accuracy of two other methods; Global Thresholding and Active Contour. Results are presented both visually and quantitatively using the Dice overlap measure. Results show that applying the Active Contour with Shape Prior method leads to more accurate segmentations than Global Thresholding and Active Contour methods. Furthermore, accuracy and robustness of the proposed method is increased if it is applied on Otsu thresholded images instead of the original grayscale versions.