ObjectiveIt remains unclear whether Tai Chi is effective for preventing falls in older adults. We undertook this systematic review to evaluate the preventive effect of Tai Chi by updating the latest trial evidence.DesignSystematic review and meta-analysis.MethodsThe Cochrane Library, MEDLINE and EMBASE were searched up to February 2016 to identify randomised trials evaluating Tai Chi for preventing falls in older adults. We evaluated the risk of bias of included trials using the Cochrane Collaboration's tool. Results were combined using random effects meta-analysis.Outcome measuresNumber of fallers and rate of falls.Results18 trials with 3824 participants were included. The Tai Chi group was associated with significantly lower chance of falling at least once (risk ratio (RR) 0.80, 95% CI 0.72 to 0.88) and rate of falls (incidence rate ratio (IRR) 0.69, 95% CI 0.60 to 0.80) than the control group. Subgroup analyses suggested that the preventive effect was likely to increase with exercise frequency (number of fallers: p=0.001; rate of falls: p=0.007) and Yang style Tai Chi was likely to be more effective than Sun style Tai Chi (number of fallers: p=0.01; rate of falls: p=0.001). The results might be influenced by publication bias as the funnel plots showed asymmetry. Sensitivity analyses by sample size, risk of bias and comorbidity showed no major influence on the primary results.ConclusionsTai Chi is effective for preventing falls in older adults. The preventive effect is likely to increase with exercise frequency and Yang style Tai Chi seems to be more effective than Sun style Tai Chi.

The diagnosis of brain tumor types generally depends on the clinical experience of doctors, and computer-assisted diagnosis improves the accuracy of diagnosing tumor types. Therefore, a convolutional neural network based on complex networks (CNNBCN) with a modified activation function for the magnetic resonance imaging classification of brain tumors is presented. The network structure is not manually designed and optimized, but is generated by randomly generated graph algorithms. These randomly generated graphs are mapped into a computable neural network by a network generator. The accuracy of the modified CNNBCN model for brain tumor classification reaches 95.49%, which is higher than several models presented by other works. In addition, the test loss of brain tumor classification of the modified CNNBCN model is lower than those of the ResNet, DenseNet and MobileNet models in the experiments. The modified CNNBCN model not only achieves satisfactory results in brain tumor image classification, but also enriches the methodology of neural network design. INDEX TERMS Convolutional neural network, complex networks, randomly generated graph, network generator, brain tumors.

Wine, an alcoholic beverage made from fermented grapes, has become an increasingly popular drink. However, wine regions may directly affect the quality and taste of the wine, and misjudgment of the wine regions leads to confusion for dealers and consumers in choosing wine types. In recent decades, different methods in machine learning have been presented and investigated the pattern classification. In this paper, based on the existing results, a modified multi-output Chebyshev-polynomial feed-forward neural network (MOCPFFNN) is presented, analyzed, and applied to the pattern classification of wines regions. According to the orthogonal polynomial theory, the activation functions of the MOCPFFNN are improved to some Chebyshev polynomials. In addition, to have a lower computational complexity, the presented neural network model is automatically determined by the eight-fold cross validation (8FCV) algorithm and the weight direct determination (WDD) algorithm. Finally, comparisons are made among the presented model and other classical methods, e.g., feed-forward back propagation (FFBP), layer recurrent neural network (LRNN), and nonlinear auto regressive with exogenous inputs (NARX), K-nearest neighbors (KNN), random forest, which confirm that the modified MOCPFFNN has the best approximation and generalization performance in the pattern classification of wine regions, with the accuracy rates of the training set and test set reaching 99.17% and 94.83%, respectively. Moreover, the variance of the accuracy of the presented MOCPFFNN method in the experiments is 0, which illustrates its high robustness in pattern classification. INDEX TERMS Eight fold cross validation (8FCV), multi-output Chebyshev-polynomial feed-forward neural network (MOCPFFNN), weight direct determination (WDD), wine region.

No abstract

No abstract

Polymer hydrogels are generally insufficient biomechanics, strong resistance to cell adhesion, and weak bioactivity which limits their application in bone tissue engineering considerably. In order to develop a bone tissue engineering material with both good mechanical properties, osteogenic and angiogenic activity. Nanofibers carrying DNA plasmid (pNF) are introduced to gelatin methacryloyl (GelMA) and thiolated chitosan (TCS) system for preparing a novel GelMA/TCS/pNF composite hydrogel with dual network structure. By characterization of the compressive measurements, the resulting composite scaffold shows greatly enhanced mechanical strength (0.53 MPa) and is not damaged after 20 cycles of compression. And the fabricated composite scaffold displays sustained release of bone morphogenetic protein‐2 that can induce osteogenic differentiation and angiopoietin‐1 that promotes vascularization. The cell experiment shows that this system can significantly promote MC3T3‐E1 cell attachment, proliferation, as well as osteogenic‐related and angiogenic‐related genes expression of MC3T3‐E1 cells. Moreover, the in vivo results show that the composite scaffold with activated gene fibers can significantly promote osteogenesis and vascularization leading to favorable capacity of bone regeneration, meaning that the resulting biomimetic composite hydrogel scaffolds are excellent candidates for bone repair materials.

Deep learning models often have complicated structures with low computational speed and the requirement of a large amount of storage space, which limits their own practical application on some devices with insufficient computing power. This paper proposes the weight and structure determination neural network aided with double pseudoinversion (WASDNN-DP) that can overcome these shortcomings. First, the model structure, theoretical bases, and the algorithms of WASDNN-DP are given. In the process of constructing the network, the weight matrix between the hidden layer and the output layer is first randomly generated. After the weight matrix between the input layer and the hidden layer is analytically determined, the weight matrix between the hidden layer and the output layer is re-determined by the pseudo-inverse method. Furthermore, in WASDNN-DP, the structure of the neural network is determined by a progressive method. Subsequently, based on two datasets collected from children aged 7-15 years by using smart insoles, the comparative experiments between WASDNN-DP and some other machine learning models are carried out, which illustrate the superiority of the proposed WASDNN-DP in the diagnosis of flat foot. INDEX TERMS Classification algorithms, feedforward neural networks, weight and structure determination (WASD), computer aided diagnosis.

Flatfoot is a common disease in children and juveniles. If the disease is not controlled and treated in time, it may last into adulthood, which can bring a great deal of inconvenience and even pain to daily life. In addition to the diagnosis of the disease simply by doctors and medical equipment, artificial intelligence has become a very promising auxiliary diagnostic tool. In this paper, a neural network with a simple structure is used to classify the foot data to achieve the function of diagnosing flatfoot. The presented neural network is termed as modified weights-and-structure-determination neural network (MWASDNN), of which the input weights are analytically determined by the pseudo-inverse method, while the output weights are randomly generated within a specified interval, and the number of hidden-layer neurons is determined by an incremental method. In addition, the stratified cross-validation method is introduced to choose the model structure that best fits the features of the data set, thereby improving the generalization performance and robustness of the MWASDNN. Utilizing the MWASDNN models to classify the foot data we collected, we finally get the accuracy of 84.31% and 85.29% on the left and right foot data, respectively. Besides, MWASDNN achieves the highest classification accuracy on our foot data set compared to some traditional neural networks, pattern classification methods, and two improved neural networks. These excellent results indicate that the MWASDNN is expected to be designed as a practical flatfoot diagnostic tool.INDEX TERMS Flatfoot diagnosis, modified weights-and-structure-determination neural network (MWASDNN), pattern classification, stratified cross-validation.