No Cover Image

Conference contribution 347 views

Graph Based Lymphatic Vessel Wall Localisation and Tracking / Ehab Essa; Jonathan Jones; Xianghua Xie

Graph-Based Representations in Pattern Recognition, Volume: 9069, Pages: 345 - 354

Swansea University Author: Xie, Xianghua

Full text not available from this repository: check for access using links below.

DOI (Published version): 10.1007/978-3-319-18224-7_34

Abstract

We present a novel hidden Markov model (HMM) based approach to segment and track the lymph vessel in confocal microscopy images. The vessel borders are parameterised by radial basis functions (RBFs) so that the number of tracking points are reduced to a very few. The proposed method tracks the hidde...

Full description

Published in: Graph-Based Representations in Pattern Recognition
Published: Springer 2015
Online Access: http://link.springer.com/chapter/10.1007%2F978-3-319-18224-7_34
URI: https://cronfa.swan.ac.uk/Record/cronfa22232
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract: We present a novel hidden Markov model (HMM) based approach to segment and track the lymph vessel in confocal microscopy images. The vessel borders are parameterised by radial basis functions (RBFs) so that the number of tracking points are reduced to a very few. The proposed method tracks the hidden states that determine the border location along a set of normal lines obtained from the previous frame. The border observation is derived from edge-based features using steerable filters. Two Gaussian probability distributions for the vessel border and background are used to infer the emission probability. The transition probability is learnt by using the Baum-Welch algorithm. A new optimisation method for determining the best sequence of the hidden states is introduced. We transform the segmentation problem into a minimisation of s-excess graph cost. Each node in the graph corresponds to one state, and the weight for each node is defined using its emission probability. The inter-relation between neighbouring nodes is defined using the transition probability. Its optimal solution can be found in polynomial time using the s-t cut algorithm. Qualitative and quantitative analysis of the method on lymphatic vessel segmentation show superior performance of the proposed method compared to the traditional Viterbi algorithm.
Keywords: Graph cut, image segmentation, tracking, medical image analysis
College: College of Science
Start Page: 345
End Page: 354