Çoklu beyin yapılarının bağlaşık, parametrik olmayan şekil önbilgisi kullanılarak bölütlenmesi = Segmentation of multiple brain structures using coupled nonparametric shape priors
Uzunbaş, Mustafa Gökhan and Çetin, Müjdat and Ünal, Gözde and Erçil, Aytül (2008) Çoklu beyin yapılarının bağlaşık, parametrik olmayan şekil önbilgisi kullanılarak bölütlenmesi = Segmentation of multiple brain structures using coupled nonparametric shape priors. In: IEEE Conference on Signal Processing, Communications, and their Applications, Aydin. Turkey
Official URL: http://www.ii.metu.edu.tr/SIU2008/kurultay_programi.php
We consider the problem of segmenting multiple brain structures in medical images. Due to the low contrast of medical images and the presence of noise, solution of this problem based only on image data, is rather challenging. Motivated by this observation, we propose an appraoch that incorporates prior information about the shapes of the anatomical structures, as well as about the interaction of neighboring shapes. We construct a statistical framework for the segmentation problem, which captures information about the shapes of coupled anatomical structures through prior probability density functions. We nonparametrically estimate these probability density functions from training shapes. We develop an active contour-based segmentation algorithm that combines image-based data with shape information. We demonstrate the benefits of our approach over existing methods through challenging segmentation scenarios on real magnetic resonance images.
Repository Staff Only: item control page