The Classification of Meningioma Subtypes Based on the Color Segmentation and Shape Features
This paper proposed an automatic method for the classification of meningioma subtypes based on the unsupervised color segmentation method and feature selection scheme. Firstly, a color segmentation method is utilized to segment the cell nuclei. Then the s
- PDF / 543,104 Bytes
- 6 Pages / 439.37 x 666.142 pts Page_size
- 58 Downloads / 160 Views
The Classification of Meningioma Subtypes Based on the Color Segmentation and Shape Features Ziming Zeng, Zeng Tong, Zhonghua Han, Yinlong Zhang and Reyer Zwiggelaar
Abstract This paper proposed an automatic method for the classification of meningioma subtypes based on the unsupervised color segmentation method and feature selection scheme. Firstly, a color segmentation method is utilized to segment the cell nuclei. Then the set of shape feature vectors which are calculated from the segmentation results are constructed. Finally, a k-nearest neighbour classifier (kNN) is used to classify the meningioma subtypes. Experiment shows that the classification accuracy of 85 % is achieved by using a leave-one-out cross validation approach on 80 meningioma images. Keywords Meningioma
Segmentation Classification Color Shape features
335.1 Introduction Meningiomas are a diverse set of tumors generated from the meninges and the membranous layers surrounding the central nervous system. According to the world health organization (WHO), the majority of benign WHO Grade I
Z. Zeng (&) Z. Han Information and Control Engineering Faculty, Shenyang Jianzhu University, Liaoning, China e-mail: [email protected] Z. Zeng R. Zwiggelaar Department of Computer Science, Aberystwyth University, Aberystwyth, UK Z. Tong School of Management, Shenyang Jianzhu University, Shenyang, Liaoning, China Z. Han Y. Zhang Shenyang Institute of Automation, Chinese Academy of Science, Shenyang, Liaoning, China
S. Li et al. (eds.), Frontier and Future Development of Information Technology 2669 in Medicine and Education, Lecture Notes in Electrical Engineering 269, DOI: 10.1007/978-94-007-7618-0_335, Springer Science+Business Media Dordrecht 2014
2670
Z. Zeng et al.
meningioma can be categorized into four subtypes: fibroblastisch, meningotheliomatoes, psammomatoes and transitional. Computerized image analysis may enable an objective, standardized, and time-saving assessment of these prognostic features. Several automatic meningioma classification methods are proposed. Qureshi et al. [1] proposed a method based on discriminant wavelet packets (DWP) and learning vector quantization (LVQ) to classify the meningioma subtypes. In [2], an adaptive discriminant wavelet packet transform and local binary patterns (LBP) is utilized for meningioma subtype classification. In [3], a multiresolution analysis technique is introduced to resolve the issue of intra-class texture variation on the basis of stability of adaptive discriminant wavelet packet transform (ADWPT). However, most methods on meningioma classification fail to consider the features of the shape of the regions of interests (ROI). In this paper, an automated classification method is presented to classify the meningioma subtypes based on an unsupervised segmentation scheme and shape features.
335.2 The Proposed Method The proposed method contains two steps. In the first step, the density inhomogeneity in the gray level images is corrected and the density of cell nuclei is enhanced
Data Loading...
