ochang Gu

A Modified Ant Colony Optimization Algorithm for Tumor Marker Gene Selection

Microarray data are often extremely asymmetric in dimensionality, such as thousands or even tens of thousands of genes but only a few hundreds of samples or less. Such extreme asymmetry between the dimensionality of genes and samples can lead to inaccurate diagnosis of disease in clinic. Therefore, it has been shown that selecting a small set of marker genes can lead to improved classification accuracy. In this paper, a simple modified ant colony optimization (ACO) algorithm is proposed to select tumor-related marker genes, and support vector machine (SVM) is used as classifier to evaluate the performance of the extracted gene subset. Experimental results on several benchmark tumor microarray datasets showed that the proposed approach produces better recognition with fewer marker genes than many other methods. It has been demonstrated that the modified ACO is a useful tool for selecting marker genes and mining high dimension data.

General Tendencies in Segmentation of Medical Ultrasound Images

Image segmentation plays an important role in both qualitative and quantitative analysis of medical ultrasound images. But the performance of the classical image-segmentation techniques degrades severely when they are applied to segment medical ultrasound images, for medical ultrasound images have features of poor contrast and strong speckle noise. Firstly, this article investigates and compiles some of the techniques mostly used in the segmentation of medical ultrasound images. Then a bibliographical survey of current research of medical ultrasound images segmentation is given in this paper. Finally, the general tendencies of medical ultrasound images segmentation are presented.

A Novel Discrete Particle Swarm Optimization Algorithm for Microarray Data-Based Tumor Marker Gene Selection

Micorarray data are often extremely asymmetric in dimensionality, such as thousands or even tens of thousands of genes and a few hundreds of samples. Such extreme asymmetry between the dimensionality of genes and samples can lead inaccurate diagnosis of disease in clinic. Therefore, it has been shown that selecting a small set of marker genes can lead to improved classification accuracy. In this paper, a novel marker gene selection approach is proposed. Firstly, some top-ranked informative genes are selected by signal-noise ratio estimation method. Then a novel discrete particle swarm optimization (PSO) algorithm is applied to select a few marker genes and support vector machines (SVM) is used as evaluator for getting better prediction performance. Experiments show that the proposed method produces better recognition with fewer marker genes than many other methods on colon tumor dataset. It has been demonstrated the modified discrete PSO is a useful tool for selecting marker genes and mining high dimension data.

Speckle Noise Suppression Techniques for Ultrasound Images

Speckle is a multiplicative noise that degrades the visual evaluation in ultrasound imaging. In medical ultrasound image processing, speckle noise suppression has become a very essential exercise for diagnose. The recent advancements in ultrasound devices necessitate the need of more robust despeckling techniques for enhancing ultrasound medical imaging in routine clinical practice. Many denoising techniques have been proposed for effective suppression of speckle noise. This paper compiles the performance of various techniques in medical B-mode ultrasound images.

Incremental Tumor Diagnosis Algorithm Using New Unlabeled Microarray

This paper presents an implementation of incremental tumor diagnosis algorithm (ITDA) on microarray data for improving diagnostic accuracy of tumor. A classifier (BP or KNN) was used in the algorithm to estimate confidences of a new unlabeled sample in different classes. When one confidence is higher than the threshold, the sample will be labeled; otherwise, the sample will be diagnosed by medical expert with other approaches. With more and more new labeled samples adding in the labeled samples list, the classifier will be better and will provide more accurate diagnosis for new sample. Strictly speaking, it is a particular active learning algorithm. By applying this algorithm on Colon tumor dataset, we demonstrated that our incremental tumor diagnosis algorithm can be used to successfully improve diagnostic accuracy of tumor. The performance of different parameters was tested in our experiments to verify the applicability of the method.

Kernel Null Foley-Sammon Transform

The proposed null Foley-Sammon transform (NFST) method based on the Gram-Schmidt orthogonalization successfully overcomes the so-called small sample size problem with high performance in terms of recognition accuracy and low computation cost, however, the NFST method is still a linear technique in nature, so a new nonlinear feature extraction method called kernel null Foley-Sammon transform (KNFST) is presented in this paper. A major advantage of the proposed method is that it is regarded every column of the kernel matrix as a corresponding sample, which is different from other commonly used kernel-based learning algorithms. Then running NFST the in kernel matrix, nonlinear features can be extracted. Experimental results on ORL database indicate that the proposed KNFST method achieves higher recognition rate than the NFST method and other kernel-based learning algorithms.

A Neural Network Based Approach for Reliability Analysis of Software-Intensive Equipment

Software-Intensive Equipment is the system which includes software and hardware. In this paper, we analyze the characteristics of software-intensive equipment and propose a non-parametric system reliability model to study the failure data with time series technique. The model uses fuzzy neural network and a wavelet function as the membership function to adjust the shape on line so that the model has better learning and adaptive ability. Experimental results demonstrate that the predictability of the proposed model is acceptable, we expect the model feasible.

A framework for microarray data-based tumor diagnostic system with improving performance incrementally

Gene expression data obtained from DNA microarrays have shown useful in tumor classification problems. However, most existing related literatures focused on how to extract tumor-related genes and design appropriate classification strategies, but neglected effect of future unlabeled samples which are expensive to label. In this paper, we propose a novel framework to construct microarray data-based tumor diagnostic system with improving performance incrementally. Through the proposed framework, system is permitted to evaluate confidences of a new unlabeled sample in each class and opportunity of misdiagnosis decreases by returning uncertain samples to medical experts. Moreover, the system is also enabled to improve predictive accuracy by learning new experiences from incremental labeled samples constantly. The proposed framework of system has been tested on two well-known tumor microarray datasets with encouraging results and shown great potential for the developments of generic platform for tumor clinical diagnosis based on microarray data.

A new AdaboostSVM algorithm based on multi-feature fusion for multi-pose face detection

To improve the performance of multi-pose face detection, the AdaboostSVM algorithm based on multi-feature fusion is proposed in this paper. Firstly, the Haar-like features and the triangular integral features are introduced and the edge-orientation field features based on morphological gradient are presented. Then, the AdaboostSVM Algorithm based on the above three kinds of features is proposed. The results of the experiment show that the proposed algorithm could improve the performance of multi-pose face detection effectively.

An Improved Algorithm for Diverse AdaBoostSVM

In order to improve the training convergence speed and detection accuracy of Diverse AdaBoostSVM, an improved algorithm is proposed according to the asymmetry in face detection. In the algorithm, the weight of each weak learner, which represents importance of each weak learner, is determined by the error rate and the recognition capability of the weak learner for the face samples. The results of the experiments show that the proposed algorithm could improve the training convergence speed and the detection accuracy in face detection.