Training ROI Selection Based on MILBoost for Liver Cirrhosis Classification Using Ultrasound Images
Ultrasound images are widely used for diagnosis of liver cirrhosis. In most of liver ultrasound images analysis, regions of interest (ROIs) are selected carefully, to use for feature extraction and classification. It is difficult to select ROIs exactly fo
- PDF / 1,820,254 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 97 Downloads / 182 Views
Abstract. Ultrasound images are widely used for diagnosis of liver cirrhosis. In most of liver ultrasound images analysis, regions of interest (ROIs) are selected carefully, to use for feature extraction and classification. It is difficult to select ROIs exactly for training classifiers, because of the low SN ratio of ultrasound images. In these analyses, training sample selection is important issue to improve classification performance. In this article, we have proposed training ROI selection using MILBoost for liver cirrhosis classification. In our experiments, the proposed method was evaluated using manually selected ROIs. Experimental results show that the proposed method improve classification performance, compared to previous method, when qualities of class label for training sample are lower. Keywords: Ultrasound imaging MILBoost Pattern recognition Computer-aided diagnosis (CAD) Sample selection
1 Introduction Computerized analysis using image processing techniques is expected to improve medical image interpretation, because it may serve as a “second opinion” in detecting lesions, assessing disease severity, and making diagnostic decisions [1]. Ultrasound imaging is a popular and non-invasive tool that is used in the diagnoses of liver disease. The purpose of our study is cirrhosis classification with high accuracy using liver ultrasound images. However, in general, regions of interest (ROIs) on liver ultrasound images are defined manually by medical doctors. Examples of liver ultrasound images of cirrhosis and healthy cases are shown in Fig. 1. It is difficult to select the ROIs in liver ultrasound images, which are used for classifier design and classification. The classification © Springer International Publishing Switzerland 2016 H. Fujita et al. (Eds.): IEA/AIE 2016, LNAI 9799, pp. 451–459, 2016. DOI: 10.1007/978-3-319-42007-3_39
452
Y. Fujita et al.
(a) Cirrhosis
(b) Healthy
Fig. 1. Examples of liver ultrasound images.
accuracy depend on how to define ROI in the liver’s tissues, because they may contain unnecessary edges or textures by interference of speckle noise. We have proposed a method of cirrhosis detection based on multiple-ROI combination [2] by the product rule [3]. Therefore, in the classification process, blood vessel exclusion and reject option were proposed to select ROIs automatically [4]. Wu et al. also used the genetic algorithm to select ROIs to improve the diagnosis accuracy [5]. In this article, training ROI selection using MILBoost [6] is introduced to improve the cirrhosis classification performance. In our experiments, the proposed method was evaluated using manually selected ROIs. Experimental results show the proposed method improve classification performance. This article is organized as follows: Sect. 2 describes the related works. Section 3 introduces the proposed method, including feature extraction using Gabor filters, classifier design using MILBoost and multiple-ROI classification. Section 4 shows experimental results, and conclusions are described in Sect. 5.
2 Related Wo
Data Loading...
