Fast Cell Segmentation Using Scalable Sparse Manifold Learning and Affine Transform-Approximated Active Contour
Efficient and effective cell segmentation of neuroendocrine tumor (NET) in whole slide scanned images is a difficult task due to a large number of cells. The weak or misleading cell boundaries also present significant challenges. In this paper, we propose
- PDF / 1,992,225 Bytes
- 8 Pages / 439.363 x 666.131 pts Page_size
- 83 Downloads / 195 Views
1 Department of Electrical and Computer Engineering J. Crayton Pruitt Family Department of Biomedical Engineering, University of Florida, Gainesville, FL 32611, USA
Abstract. Efficient and effective cell segmentation of neuroendocrine tumor (NET) in whole slide scanned images is a difficult task due to a large number of cells. The weak or misleading cell boundaries also present significant challenges. In this paper, we propose a fast, high throughput cell segmentation algorithm by combining top-down shape models and bottom-up image appearance information. A scalable sparse manifold learning method is proposed to model multiple subpopulations of different cell shape priors. Followed by a shape clustering on the manifold, a novel affine transform-approximated active contour model is derived to deform contours without solving a large amount of computationallyexpensive Euler-Lagrange equations, and thus dramatically reduces the computational time. To the best of our knowledge, this is the first report of a high throughput cell segmentation algorithm for whole slide scanned pathology specimens using manifold learning to accelerate active contour models. The proposed approach is tested using 12 NET images, and the comparative experiments with the state of the arts demonstrate its superior performance in terms of both efficiency and effectiveness.
1
Introduction
Effective and efficient cell segmentation of pancreatic neuroendocrine tumor (NET) is a prerequisite for quantitative image analyses such as Ki67 counting. Many state-of-the-art approaches [11,4,16,10] have been applied to cell/nucleus segmentation on specific medical images. In order to handle partial occlusion, shape prior models have been introduced to improve touching cell separation [2,14] and liver segmentation [17]. However, it is inefficient to exploit the aforementioned shape prior models, which are not adaptive to large data sets, to fast segment thousands of cells in whole slide scanned specimens. In addition, it is necessary to learn multiple subpopulations of shape priors to handle shape variations. In this paper, we propose a high throughput and large-scale cell segmentation algorithm by combing high-level shape priors and low-level active contour models. The main contributions are: 1) A scalable sparse manifold learning algorithm to model multiple cell shape priors; 2) A novel affine transform-approximated active contour model that dramatically accelerates the shape deformation. c Springer International Publishing Switzerland 2015 N. Navab et al. (Eds.): MICCAI 2015, Part III, LNCS 9351, pp. 332–339, 2015. DOI: 10.1007/978-3-319-24574-4_40
Fast Cell Segmentation
2
333
Methodology
Singular Values
4 An effective cell segmentation framex 10 8 work combining shape prior models and image appearance information is 6 presented in [14]; however, it requires to solve one associated partial dif4 ferential equation for each contour within each iteration and therefore is 2 not suitable to handle a large number of cells in whole slide scanned 0 0 10 20 30 images. In this p
Data Loading...