Suxian Cai

Knee Joint Vibration Signal Analysis with Matching Pursuit Decomposition and Dynamic Weighted Classifier Fusion

Analysis of knee joint vibration (VAG) signals can provide quantitative indices for detection of knee joint pathology at an early stage. In addition to the statistical features developed in the related previous studies, we extracted two separable features, that is, the number of atoms derived from the wavelet matching pursuit decomposition and the number of significant signal turns detected with the fixed threshold in the time domain. To perform a better classification over the data set of 89 VAG signals, we applied a novel classifier fusion system based on the dynamic weighted fusion (DWF) method to ameliorate the classification performance. For comparison, a single leastsquares support vector machine (LS-SVM) and the Bagging ensemble were used for the classification task as well. The results in terms of overall accuracy in percentage and area under the receiver operating characteristic curve obtained with the DWF-based classifier fusion method reached 88.76% and 0.9515, respectively, which demonstrated the effectiveness and superiority of the DWF method with two distinct features for the VAG signal analysis.

Representation of fluctuation features in pathological knee joint vibroarthrographic signals using kernel density modeling method

This article applies advanced signal processing and computational methods to study the subtle fluctuations in knee joint vibroarthrographic (VAG) signals. Two new features are extracted to characterize the fluctuations of VAG signals. The fractal scaling index parameter is computed using the detrended fluctuation analysis algorithm to describe the fluctuations associated with intrinsic correlations in the VAG signal. The averaged envelope amplitude feature measures the difference between the upper and lower envelopes averaged over an entire VAG signal. Statistical analysis with the Kolmogorov-Smirnov test indicates that both of the fractal scaling index (p=0.0001) and averaged envelope amplitude (p=0.0001) features are significantly different between the normal and pathological signal groups. The bivariate Gaussian kernels are utilized for modeling the densities of normal and pathological signals in the two-dimensional feature space. Based on the feature densities estimated, the Bayesian decision rule makes better signal classifications than the least-squares support vector machine, with the overall classification accuracy of 88% and the area of 0.957 under the receiver operating characteristic (ROC) curve. Such VAG signal classification results are better than those reported in the state-of-the-art literature. The fluctuation features of VAG signals developed in the present study can provide useful information on the pathological conditions of degenerative knee joints. Classification results demonstrate the effectiveness of the kernel feature density modeling method for computer-aided VAG signal analysis.

Effective Dysphonia Detection Using Feature Dimension Reduction and Kernel Density Estimation for Patients with Parkinson's Disease

Detection of dysphonia is useful for monitoring the progression of phonatory impairment for patients with Parkinson’s disease (PD), and also helps assess the disease severity. This paper describes the statistical pattern analysis methods to study different vocal measurements of sustained phonations. The feature dimension reduction procedure was implemented by using the sequential forward selection (SFS) and kernel principal component analysis (KPCA) methods. Four selected vocal measures were projected by the KPCA onto the bivariate feature space, in which the class-conditional feature densities can be approximated with the nonparametric kernel density estimation technique. In the vocal pattern classification experiments, Fisher’s linear discriminant analysis (FLDA) was applied to perform the linear classification of voice records for healthy control subjects and PD patients, and the maximum a posteriori (MAP) decision rule and support vector machine (SVM) with radial basis function kernels were employed for the nonlinear classification tasks. Based on the KPCA-mapped feature densities, the MAP classifier successfully distinguished 91.8% voice records, with a sensitivity rate of 0.986, a specificity rate of 0.708, and an area value of 0.94 under the receiver operating characteristic (ROC) curve. The diagnostic performance provided by the MAP classifier was superior to those of the FLDA and SVM classifiers. In addition, the classification results indicated that gender is insensitive to dysphonia detection, and the sustained phonations of PD patients with minimal functional disability are more difficult to be correctly identified.

Classification of Knee Joint Vibration Signals Using Bivariate Feature Distribution Estimation and Maximal Posterior Probability Decision Criterion

Analysis of knee joint vibration or vibroarthrographic (VAG) signals using signal processing and machine learning algorithms possesses high potential for the noninvasive detection of articular cartilage degeneration, which may reduce unnecessary exploratory surgery. Feature representation of knee joint VAG signals helps characterize the pathological condition of degenerative articular cartilages in the knee. This paper used the kernel-based probability density estimation method to model the distributions of the VAG signals recorded from healthy subjects and patients with knee joint disorders. The estimated densities of the VAG signals showed explicit distributions of the normal and abnormal signal groups, along with the corresponding contours in the bivariate feature space. The signal classifications were performed by using the Fisher’s linear discriminant analysis, support vector machine with polynomial kernels, and the maximal posterior probability decision criterion. The maximal posterior probability decision criterion was able to provide the total classification accuracy of 86.67% and the area (Az) of 0.9096 under the receiver operating characteristics curve, which were superior to the results obtained by either the Fisher’s linear discriminant analysis (accuracy: 81.33%, Az: 0.8564) or the support vector machine with polynomial kernels (accuracy: 81.33%, Az: 0.8533). Such results demonstrated the merits of the bivariate feature distribution estimation and the superiority of the maximal posterior probability decision criterion for analysis of knee joint VAG signals.

Removal of artifacts in knee joint vibroarthrographic signals using ensemble empirical mode decomposition and detrended fluctuation analysis.

High-resolution knee joint vibroarthrographic (VAG) signals can help physicians accurately evaluate the pathological condition of a degenerative knee joint, in order to prevent unnecessary exploratory surgery. Artifact cancellation is vital to preserve the quality of VAG signals prior to further computer-aided analysis. This paper describes a novel method that effectively utilizes ensemble empirical mode decomposition (EEMD) and detrended fluctuation analysis (DFA) algorithms for the removal of baseline wander and white noise in VAG signal processing. The EEMD method first successively decomposes the raw VAG signal into a set of intrinsic mode functions (IMFs) with fast and low oscillations, until the monotonic baseline wander remains in the last residue. Then, the DFA algorithm is applied to compute the fractal scaling index parameter for each IMF, in order to identify the anti-correlation and the long-range correlation components. Next, the DFA algorithm can be used to identify the anti-correlated and the long-range correlated IMFs, which assists in reconstructing the artifact-reduced VAG signals. Our experimental results showed that the combination of EEMD and DFA algorithms was able to provide averaged signal-to-noise ratio (SNR) values of 20.52 dB (standard deviation: 1.14 dB) and 20.87 dB (standard deviation: 1.89 dB) for 45 normal signals in healthy subjects and 20 pathological signals in symptomatic patients, respectively. The combination of EEMD and DFA algorithms can ameliorate the quality of VAG signals with great SNR improvements over the raw signal, and the results were also superior to those achieved by wavelet matching pursuit decomposition and time-delay neural filter.

Adaptive Linear and Normalized Combination of Radial Basis Function Networks for Function Approximation and Regression

This paper presents a novel adaptive linear and normalized combination (ALNC) method that can be used to combine the component radial basis function networks (RBFNs) to implement better function approximation and regression tasks. The optimization of the fusion weights is obtained by solving a constrained quadratic programming problem. According to the instantaneous errors generated by the component RBFNs, the ALNC is able to perform the selective ensemble of multiple leaners by adaptively adjusting the fusion weights from one instance to another. The results of the experiments on eight synthetic function approximation and six benchmark regression data sets show that the ALNC method can effectively help the ensemble system achieve a higher accuracy (measured in terms of mean-squared error) and the better fidelity (characterized by normalized correlation coefficient) of approximation, in relation to the popular simple average, weighted average, and the Bagging methods.

Chondromalacia Patellae Detection by Analysis of Intrinsic Mode Functions in Knee Joint Vibration Signals

Conference Name:World Congress on Medical Physics and Biomedical Engineering. Conference Address: Beijing, China. Time:May 26, 2012 - May 31, 2012.

Statistical Analysis of Gait Maturation in Children Using Nonparametric Probability Density Function Modeling

Analysis of gait dynamics in children may help understand the development of neuromuscular control and maturation of locomotor function. This paper applied the nonparametric Parzen-window estimation method to establish the probability density function (PDF) models for the stride interval time series of 50 children (25 boys and 25 girls). Four statistical parameters, in terms of averaged stride interval (ASI), variation of stride interval (VSI), PDF skewness (SK), and PDF kurtosis (KU), were computed with the Parzen-window PDFs to study the maturation of stride interval in children. By analyzing the results of the children in three age groups (aged 3–5 years, 6–8 years, and 10–14 years), we summarize the key findings of the present study as follows. (1) The gait cycle duration, in terms of ASI, increases until 14 years of age. On the other hand, the gait variability, in terms of VSI, decreases rapidly until 8 years of age, and then continues to decrease at a slower rate. (2) The SK values of both the histograms and Parzen-window PDFs for all of the three age groups are positive, which indicates an imbalance in the stride interval distribution within an age group. However, such an imbalance would be meliorated when the children grow up. (3) The KU values of both the histograms and Parzen-window PDFs decrease with the body growth in children, which suggests that the musculoskeletal growth enables the children to modulate a gait cadence with ease. (4) The SK and KU results also demonstrate the superiority of the Parzen-window PDF estimation method to the Gaussian distribution modeling, for the study of gait maturation in children.

Classification of knee joint vibroarthrographic signals using k-nearest neighbor algorithm

The pathological condition in a degenerative knee joint may be assessed by analyzing the knee joint vibroarthrographic signals. With the severity level of the knee joint disorders evaluated by the computational methods, unnecessary imaging examination or open surgery can be prevented. In the present study, we used the k-nearest neighbor (k-NN) algorithm, a type of lazy learning approach, to classify the knee joint vibroarthrographic signals collected from healthy subjects and symptomatic patients with knee joint disorders. With the representative features of form factor and variance of the mean-square values, the k-NN algorithm is able to correctly discriminate 80% signals with the sensitivity of 0.71 and the specificity of 0.85, which is superior to the total accurate rate of 77% (sensitivity: 0.64, specificity: 0.85) provided by the Fishers linear discriminant analysis.

An artificial-neural-network-based multiple classifier system for knee-joint vibration signal classification

The knee-joint vibration or vibroarthrographic (VAG) signal could be used as an indicator with regard to the condition of degenerative articular cartilage surfaces of the knee joint. Analysis of VAG signals can assist in the screening for knee-joint pathology and help prevent unnecessary exploratory surgery. This paper proposes a multiple classifier system (MCS) based on artificial neural networks for the classification of VAG signals with statistical features. The multiple classifier system combines a group of component least-squares support vector machine classifiers with a linear and normalized fusion model. The fusion model minimizes the mean-squared error (MSE) of the MCS by solving the corresponding constrained quadratic programming problem, and the optimal weights are derived from the energy convergence process of a recurrent neural network. The results obtained with a data set of 89 VAG signals show that the proposed MCS can effectively reduce the classification error in terms of MSE. In addition, the proposed MCS also provides an area of 0.8230 under the receiver operating characteristics curve, which is much better in comparison with any one of the component networks with different input features, and also superior to the popular simple average or weighted average fusion method.