다중혈관 관상동맥 환자에서 y-문합을 이용하여 양쪽 내흉동맥만을 사용한 우회술의 조기 성적

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

CA : A Cancer Journal for Clinicians

We conduct an exhaustive survey of image thresholding methods, categorize them, express their formulas under a uniform notation, and finally carry their performance comparison. The thresholding methods are categorized according to the information they are exploiting, such as histogram shape, measurement space clustering, entropy, object attributes, spatial correlation, and local gray-level surface. 40 selected thresholding methods from various categories are compared in the context of nondestructive testing applications as well as for document images. The comparison is based on the combined performance measures. We identify the thresholding algorithms that perform uniformly better over nonde- structive testing and document image applications. © 2004 SPIE and IS&T. (DOI: 10.1117/1.1631316)

/pdf/survey-over-image-thresholding-techniques-and-quantitative-2ekadyqjhj.pdf

Survey over image thresholding techniques and quantitative performance evaluation

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

A parallel VLSI-oriented algorithm for grading tubules in digitized images of microscopic slide information of breast cancer malignancies is presented. The techniques used are to select the bright cores of tubular areas using a single threshold and then to extend these bright cores according to the surroundings of these areas. This is effective in detecting regions that warrant further analysis. Based on the different features between tubule areas and rough texture areas (i.e. a tubule area is circular, homogeneous, and its boundary is dark), a 3*3 mask Laplacian and extension operation are used to effectively separate the tubule areas from the rough texture areas. The experimental results are satisfactory. The essential parallelism of the proposed algorithm makes it easy to implement by using a VLSI architecture, where each processing element only needs to perform addition, subtraction, or comparison. >

A parallel approach to tubule grading in breast cancer lesions and its VLSI implementation

Because of ultrasound images' low quality, fully automated segmentation of breast ultrasound (BUS) image is a challenging task. In this paper, a novel segmentation method for BUS images which is fully automatic without any human intervention is proposed. By incorporating empirical knowledge and characteristics of breast structure, a ROI is generated automatically. Then two newly proposed lesion features: phase in max-energy orientation (PMO) and radial distance (RD), combined with the commonly used intensity and texture feature, are extracted. Then the new feature set is used to distinguish lesion region from the background by a trained ANN. The proposed segmentation method was tested on a BUS database composed of 60 cases. We use the manually outlined lesions by an experienced radiologist as the golden standard and evaluated the performance by both area error metrics and boundary error metrics. Quantitative results demonstrate the efficiency of the proposed fully automatic BUS segmentation method.

Completely automatic segmentation for breast ultrasound using multiple-domain features

Breast cancer occurs to over 8% women during their lifetime, and is a leading cause of death among women. Sonography is superior to mammography in its ability to detect focal abnormalities in the dense breasts and has no side-effect. In this paper, we proposed a novel automatic segmentation algorithm based on the characteristics of breast tissue and the eliminating particle swarm optimization (EPSO) clustering analysis. The characteristics of mammary gland in breast ultrasound (BUS) images are analyzed and utilized, and a method based on step-down threshold technique is employed to locate the mammary gland area. The EPSO clustering algorithm employes the idea of “survival of the superior and weeding out the inferior”. The experimental results demonstrate that the proposed approach can segment BUS image with high accuracy and low computational time. The EPSO clustering method reduces the computational time by 32.75% compared with the standard PSO clustering algorithm. The proposed approach would find wide applications in automatic lesion classification and computer aided diagnosis (CAD) systems of breast cancer.

/pdf/a-novel-approach-to-breast-ultrasound-image-segmentation-4aln30jilj.pdf

A Novel Approach to Breast Ultrasound Image Segmentation Based on the Characteristics of Breast Tissue and Particle Swarm Optimization

Automatic tumor segmentation of breast ultrasound (BUS) image is quite challenging due to the complicated anatomic structure of breast and poor image quality. Most tumor segmentation approaches achieve good performance on BUS images collected in controlled settings; however, the performance degrades greatly with BUS images from different sources. Tumor saliency estimation (TSE) has attracted increasing attention to solve the problem by modeling radiologists' attention mechanism. In this paper, we propose a novel hybrid framework for TSE, which integrates both high-level domain-knowledge and robust low-level saliency assumptions and can overcome drawbacks caused by direct mapping in traditional TSE approaches. The new framework integrated the Neutro-Connectedness (NC) map, the adaptive-center, the correlation and the layer structure-based weighted map. The experimental results demonstrate that the proposed approach outperforms state-of-the-art TSE methods.

A Hybrid Framework for Tumor Saliency Estimation

Neutrosophic logic is a relatively new logic that is a generalization of fuzzy logic. In this paper, for the first time, neutrosophic logic is applied to the field of classifiers where a support vector machine (SVM) is adopted as the example to validate its feasibility and effectiveness. The proposed neutrosophic set is integrated into a reformulated SVM, and the performance of the obtained classifier N-SVM is evaluated under a region-based image categorization system. Images are first segmented by a hierarchical two-stage self-organizing map (HSOM), using color and texture features. A novel approach is proposed to select the training samples of HSOM based on homogeneity properties. A diverse density support vector machine (DD-SVM) framework is then applied to viewing an image as a bag of instances corresponding to the regions obtained from image segmentation. Each bag is mapped to a point in the new bag space, and the categorization is transformed to a classification problem. Then, the proposed N-SVM is used as the classifier in the new bag space. N-SVM treats samples differently according to the weighting function, and it helps to reduce the effects of outliers. Experimental results have demonstrated the validity and effectiveness of the proposed method which may find wide applications in the related areas.

Heng-Da Cheng

Papers

A parallel approach to tubule grading in breast cancer lesions and its VLSI implementation

Completely automatic segmentation for breast ultrasound using multiple-domain features

A Novel Approach to Breast Ultrasound Image Segmentation Based on the Characteristics of Breast Tissue and Particle Swarm Optimization

A Hybrid Framework for Tumor Saliency Estimation

A Novel Neutrosophic Logic SVM (N-SVM) and Its Application to Image Categorization