Search

Multi-sensor recordings are normally used in event-related potential (ERP)-based brain computer interfaces (BCIs), for capturing brain activities widely distributed over the cortical surface. However, this may lead to an increased number of sensors for boosting classification performance, as well as a complicated computational effort for optimizing/reducing sensors, limiting the popularization of mobile/wearable BCIs for the end use. The localization of brain activities may help fix this issue by making useful information concentrated on relatively local brain areas, thus greatly reducing the number of sensors required and computational burden arising from the sensor selection. In the present study, we examined localization of brain activities for an ERP speller, by using novel visual graphic stimuli to induce specific brain responses. Participants were instructed to perform a spelling task under both the graphic stimuli-based and traditional character-flashing-based ERP speller paradigms. Experimental results showed that, compared to character-flashing stimuli, localized brain activities, concentrated over the posterior region, were observed for the graphic stimuli. Classification accuracies and information transfer rates were further evaluated and compared among full- (FS), normal- (NS), and localized- (LS) sensor settings. Effects of PARADIGM, SENSORSETTING, and TRIAL LENGTH were examined by a three-way repeated measure analysis of variance (ANOVA). ANOVA results showed that, the graphic paradigm achieved significantly better performance under LS than those achieved by the traditional paradigm at any of the three sensor settings, indicating that with visual graphic stimuli, localized posterior activities were enough to drive an ERP-based speller to achieve comparable or even better performance, compared to the traditional paradigm using global activities.

It is still a challenging task to segment ultrasound medical images automatically and accurately because of the poor quality of images. To address these problems, a two-stage automatic segmentation scheme based on local entropy and a proposed active contour model for ultrasound images are put forward in this paper. First, a new region-based active contour model in the level set formulation, driven by global and local intensity information, is established for the segmentation. Furthermore, for automatically segmenting the ultrasound images, a coarse segmentation is performed utilizing local entropy information of the ultrasound images, then the coarse segmentation result is used as the initial value of the explicit segmentation based on the proposed model. Several experiments on real ultrasound images demonstrate that the proposed method outperforms other methods on both visual perception and objective evaluation metrics.

Due to the abundant noise, blurry boundaries, and intensity inhomogeneities present in ultrasound (US) images, it is a difficult task to segment US images accurately. In this paper, we propose a novel active contour method that combines global and local region information to achieve this task. The global information can segment US images with noise and blurry boundaries; while the local information can settle the intensity homogeneities of images. The proposed method can be directly applied to synthetic, real, and US-images segmentation. Results demonstrate the superiority of the proposed method over other representative algorithms. Moreover, we also extend the proposed method to vector-valued images. Experiments are performed to testify the feasibility of the method, and the proposed vector-valued idea can be applied to the medical co-segmentation in the future.

Segmentation is generally an important prior process to automatic analysis of heart sounds. The proposed algorithm achieved this in separated cycles. First, the cyclostationary envelopes of heart sound signals are calculated. Second, a threshold is chosen to separate S1 and S2. The segmentation is automatic, and robust to noise. It can identify S1 and S2 correctly even under the circumstance of noise and interference. No reference signal is needed for this segmentation. It was tested by the heart sounds of 20 subjects including 15 normal and 5 abnormal in various clinical cases (715 cycles in total). The statistics showed over 96 percent correctness for segmentation.

Due to the three dimensional motion characteristics of the target, the received scattering signal often contains a cubic term in its phase function. The existing signal model is not appropriate to approximate parameters in this case. In this paper, we propose a new signal model to accurately estimate parameters of the target. We use a novel approach, the fractional ambiguity function, to estimate Doppler frequency parameters in the fractional Fourier transform domain. Furthermore, we also develop two sub-array models to accurately estimate the direction-of-departure and direction-of-arrival by employing the proposed fractional cross-ambiguity function based MUSIC (FCAF-MUSIC) algorithm and the fractional cross-ambiguity function based ESPRIT (FCAF-ESPRIT) algorithm. Simulation results are presented to verify the effectiveness of the proposed method.

This paper presents a robust time delay estimation algorithm for the α-stable noise based on correntropy. Many time delay estimation algorithms derived for impulsive stable noise are based on the theory of Fractional Lower Order Statistics (FLOS). Unlike previously introduced FLOS-type algorithms, the new algorithm is proposed to estimate the time delay by maximizing the generalized correlation function of two observed signals needing neither prior information nor estimation of the numerical value of the stable noise’s characteristic exponent. An interval for kernel selection is found for a wide range of characteristic exponent values of α-stable distribution. Simulations show the proposed algorithm offers superior performance over the existing covariation time delay estimation, least mean p-norm time delay estimation and achieves slightly improved performance than fractional lower order covariance time delay estimation at lower signal to noise ratio when the noise is highly impulsive

An improved mixed cryptosystem is presented by combining both random phase encoding technique based chaos and public key cryptography. The scheme provides not only good robustness of random phase encoding technique, but also a new solution of key distribution for symmetric cryptography algorithms. In additions, computer simulation is illustrated in detail. The tolerance to data loss of the encoded barcode is also studied particularly. The results show that the presented encoding method has advantage of robustness and high security, and very convenient to be popularized in practice.

Many signals encountered in wireless communication, telemetry, radar and sonar systems exhibit a cyclostationarity (or spectral correlation) property, as a result of periodically carrier modulating, multiplexing and coding operations. By exploiting cyclostationarity, i.e., evaluating the cyclic correlation or cyclic spectrum of the received signals at certain cycle frequency, we can extract the measures of the signals with different cycle frequency. Thus, the signal processing algorithms which based on cyclic statistics are highly tolerant to the interference and noise. However, most of the cyclostationarity based methods assume that the noise is Gaussian noise. Hence, their performance can be degraded in real applications where the noises contain non-Gaussian impulsive components. In this paper, we propose a new approach for exploiting cyclostationarity of signals in alpha-stable impulsive noise. Simulation results indicate a significantly better performance for the new statistics in some environments.

In some communication systems the transmitted signal is contaminated by impulsive noise with a non-Gaussian distribution. Non-Gaussian noise caused significant performance degradation to communication receivers. The new constant modulus blind equalizer based on fractional lower-order statistics of the equalizer input, referred to as FLOS_CMA, is able to mitigate impulsive channel noise while restoring the constant modulus character of the transmitted communication signal. However, like the constant modulus algorithm (CMA), the steady-state mean square error of the FLOS_CMA algorithm may not be sufficiently low for the system to obtain adequate performance. This paper proposes a concurrent equalizer, in which a decision-directed Least Mean p Norm (DD_LMP) equalizer operates cooperatively with a FLOS_CMA equalizer, controlled through a nonlinear link. Simulation results using M-QAM signaling have shown that the concurrent FLOS_CMA and DD_LMP blind equalizer achieves a dramatic improvement in equalization performance over the FLOS_CMA approach.

The constant modulus algorithm based on fractional lower-order statistics (FLOS_CMA) has been proven to be an effective blind equalization method under α -stable impulsive noise environment. But, there have been little results in the literature about the performance of this algorithm. In this paper, the steady-state mean- square error (MSE) performance of the FLOS_CMA is studied, and the approximate analytical expressions for real- and complex-valued data under stationary and non- stationary environments are derived, respectively, based on the energy-preserving relation and Taylor series expansion. Finally, the simulation studies are undertaken to support the analysis.

Gadolinium-enhanced multi-phase dynamic imaging has improved the accuracy of the diagnosis of hypervascular hepatocellular carcinoma (HCC), but using gadolinium-enhanced dynamic imaging alone is problematic in evaluating hypovascular HCC. This work aimed at evaluating the combined use of superparamagnetic iron oxide (SPIO)-enhanced and gadolinium set in distinguishing HCCs from regenerative nodules (RNs) in a rat model induced by diethylnitrosamine (DEN).DEN-induced HCC model rats (n=40) and control rats (n=10) were studied. From weeks 16 to 19 after DEN administration, 4 animals were scanned every week. The hepatic changes were tested with a 1.5 Tesla magnet, and MR images of SPIO-enhanced and gadolinium set were obtained. According to the pathologic changes, the tumorigenesis was divided into HCC and RN (diameter of nodulesor =3 mm). Diagnostic accuracy of the combined SPIO-enhanced and gadolinium set and the gadolinium set alone was evaluated using receiver-operating characteristic curves. Sensitivity and specificity of the combined SPIO-enhanced and gadolinium set and the gadolinium set alone were calculated.The listed tests were completed in 29 rats (21 treated and 8 controls). One hundred and six nodules (82 HCCs, 24 RNs) were analyzed. The Az value and sensitivity with the combined SPIO-enhanced and gadolinium set (Az 0.94, sensitivity 0.96) were higher than those with the gadolinium set alone (Az 0.92, sensitivity 0.89). Using the combined SPIO-enhanced and gadolinium set led to detection of 6 nodules which were negative in the gadolinium set alone and 3 nodules were correctly characterized.Using the combined SPIO-enhanced and gadolinium set improved the detectability of HCCs and the SPIO-enhanced imaging compensated for the gadolinium set in differentiating HCCs from RNs in a rat model.

Although algorithms based on rough set (RS) theory can extract useful decision rules with the effectiveness in dealing with inexact, uncertain or vague information, the deterministic mechanism for the description of error is very simple and the rules generated by RS are often unstable and have low classification accuracy. Neural networks (NN) are considered the most powerful classifier for their low classification error rates and robustness to noise. But NN usually require long time to train the huge amount of data of large databases and lack explanation facilities for their knowledge. Therefore, we combine RS and NN for autonomous decision-making, with high accuracy, robustness to noise, efficiency, and good understandability. First, generate the decision rules based on RS, then construct the NN with the hidden layer representing decision rules, and learn the arguments of the NN with BP algorithm. With the direction of RS, NN needn’t long time for training, and the knowledge buried in their structures and weights can be well explained by decision rule on RS. The proposed algorithm has been tested on a medical data set for patients with solitary pulmonary nodules (SPN).

In this paper, a computer-aided diagnostic (CAD) system for the classification of rat liver lesions from MR imaging is presented. The proposed system consists of two modules: the feature extraction and the classification modules. 40 rats are used for hepatocellular carcinoma (HCC) induction with Diethylnitrosamine via drinking water. After Resovist is administrated by tail vein the animals are scanned by a 1.5-T MR scanner with T2-weighted FRFSE sequence. SPIO-enhanced images of 106 nodules (RNs(:) 24, HCCs: 82) are acquired, and 161 regions of interest (ROIs) are taken from the MR images .Six parameters of texture characteristics including Angular Second Moment, Contrast, Correlation, Inverse Difference Moment, Entropy, and Variance of 161 ROIs are calculated and assessed by gray-level co-occurrence matrices, then fed into a BP neural network (NN) classifier to classify the liver tissue into two classes: cirrhosis and HCC. Difference of each texture parameter between cirrhosis and HCC group is significant. The accuracy of classification of HCC nodules from cirrhosis is 91.67%. It indicates the ANN classifier based on texture is effective for classifying HCC nodules from cirrhosis on rat SPIO-enhanced imaging.