Segmentation of endothelial cell boundaries of rabbit aortic images using a machine learning approach

Authors:
Saadia Iftikhar;Andrew R. Bond;Asim I. Wagan;Peter D. Weinberg;Anil A. Bharath
Affiliations:
Department of Bioengineering, Imperial College London, London, UK;Department of Bioengineering, Imperial College London, London, UK and Bristol Heart Institute, Bristol Royal Infirmary, Bristol, UK;Laboratoire d'InfoRmatique en Image et Systèmes d'Information, Institut National des Sciences Appliquées de Lyon, LIRIS INSA De Lyon, Villeurbanne, France;Department of Bioengineering, Imperial College London, London, UK;Department of Bioengineering, Imperial College London, London, UK
Venue:
Journal of Biomedical Imaging
Year:
2011

Citing 10
Cited 0

A Computational Approach to Edge Detection

IEEE Transactions on Pattern Analysis and Machine Intelligence
Minimum error thresholding

Pattern Recognition
An introduction to support Vector Machines: and other kernel-based learning methods

An introduction to support Vector Machines: and other kernel-based learning methods
A Tutorial on Support Vector Machines for Pattern Recognition

Data Mining and Knowledge Discovery
Segmentation of ultrasonic images using support vector machines

Pattern Recognition Letters - Speciqal issue: Ultrasonic image processing and analysis
Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Feature Selection Based on Mutual Information: Criteria of Max-Dependency, Max-Relevance, and Min-Redundancy

IEEE Transactions on Pattern Analysis and Machine Intelligence
Invited talk: Can learning kernels help performance?

ICML '09 Proceedings of the 26th Annual International Conference on Machine Learning
LIBSVM: A library for support vector machines

ACM Transactions on Intelligent Systems and Technology (TIST)
A steerable complex wavelet construction and its application to image denoising

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an automatic detection method for thin boundaries of silver-stained endothelial cells (ECs) imaged using light microscopy of endotheliummono-layers from rabbit aortas. To achieve this, a segmentation technique was developed, which relies on a rich feature space to describe the spatial neighbourhood of each pixel and employs a Support Vector Machine (SVM) as a classifier. This segmentation approach is compared, using hand-labelled data, to a number of standard segmentation/thresholding methods commonly applied in microscopy. The importance of different features is also assessed using the method of minimum Redundancy, Maximum Relevance (mRMR), and the effect of different SVM kernels is also considered. The results show that the approach suggested in this paper attains much greater accuracy than standard techniques; in our comparisons with manually labelled data, our proposed technique is able to identify boundary pixels to an accuracy of 93%. More significantly, out of a set of 56 regions of image data, 43 regions were binarised to a useful level of accuracy. The results obtained from the image segmentation technique developed here may be used for the study of shape and alignment of ECs, and hence patterns of blood flow, around arterial branches.