Search

Existing algorithms of speech-based deception detection are severely restricted by the lack of sufficient number of labelled data. However, a large amount of easily available unlabelled data has not been utilized in reality. To solve this problem, this paper proposes a semi-supervised additive noise autoencoder model for deception detection. This model updates and optimizes the semi-supervised autoencoder and it consists of two layers of encoder and decoder, and a classifier. Firstly, it changes the activation function of the hidden layer in network according to the characteristics of the deception speech. Secondly, in order to prevent over-fitting during training, the specific ratio dropout is added at each layer cautiously. Finally, we directly connected the supervised classification task in the output of encoder to make the network more concise and efficient. Using the feature set specified by the INTERSPEECH 2009 Emotion Challenge, the experimental results on Columbia-SRI-Colorado (CSC) corpus and our own deception corpus show that the proposed model can achieve more advanced performance than other alternative methods with only a small amount of labelled data. [ABSTRACT FROM AUTHOR]

The prediction of cell-lines sensitivity to a given set of compounds is a very important factor in the optimization of in-vitro assays. To date, the most common prediction strategies are based upon machine learning or other quantitative structure-activity relationships (QSAR) based approaches. In the present research, we propose and discuss a straightforward strategy not based on any learning modelling but exclusively relying upon the chemical similarity of a query compound to reference compounds with annotated activity against cell lines. We also compare the performance of the proposed method to machine learning predictions on the same problem. A curated database of compounds-cell lines associations derived from ChemBL version 22 was created for algorithm construction and cross-validation. Validation was done using 10-fold cross-validation and testing the models on new data obtained from ChemBL version 25. In terms of accuracy, both methods perform similarly with values around 0.65 across 750 cell lines in 10-fold cross-validation experiments. By combining both methods it is possible to achieve 66% of correct classification rate in more than 26000 newly reported interactions comprising 11000 new compounds. A Web Service implementing the described approaches (both similarity and machine learning based models) is freely available at: . [ABSTRACT FROM AUTHOR]

Sensor network intrusion detection has attracted extensive attention. However, previous intrusion detection methods face the highly imbalanced attack class distribution problem, and they may not achieve a satisfactory performance. To solve this problem, we propose a new intrusion detection algorithm based on normalized cut spectral clustering for sensor network in this paper. The main aim is to reduce the imbalance degree among classes in an intrusion detection system. First, we design a normalized cut spectral clustering to reduce the imbalance degree between every two classes in the intrusion detection data set. Second, we train a network intrusion detection classifier on the new data set. Finally, we do extensive experiments and analyze the experimental results in detail. Simulation experiments show that our algorithm can reduce the imbalance degree among classes and reserves the distribution of the original data on the one hand, and improve effectively the detection performance on the other hand. [ABSTRACT FROM AUTHOR]

Background: The Sequential Organ Failure Assessment (SOFA) score is commonly used in ICUs around the world, designed to assess the severity of the patient's clinical state based on function/dysfunction of six major organ systems. The goal of this work is to build a computational model to predict mortality based on a series of SOFA scores. In addition, we examined the possibility of improving the prediction by incorporating a new component designed to measure the performance of the gastrointestinal system, added to the other six components. Methods: In this retrospective study, we used patients’ three latest SOFA scores recorded during an individual ICU stay as input to different machine learning models and ensemble learning models. We added three validated parameters representing gastrointestinal failure. Among others, we used classification models such as Support Vector Machines (SVMs), Neural Networks, Logistic Regression and a penalty function used to increase model robustness in regard to certain extreme cases, which may be found in ICU population. We used the Area under Curve (AUC) performance metric to examine performance. Results: We found an ensemble model of linear and logistic regression achieves a higher AUC compared related works in past years. After incorporating the gastrointestinal failure score along with the penalty function, our best performing ensemble model resulted in an additional improvement in terms of AUC metrics. We implemented and compared 36 different models that were built using both the information from the SOFA score as well as that of the gastrointestinal system. All compared models have approximately similar and relatively large AUC (between 0.8645 and 0.9146) with the best results are achieved by incorporating the gastrointestinal parameters into the prediction models. Conclusions: Our findings indicate that gastrointestinal parameters carry significant information as a mortality predictor in addition to the conventional SOFA score. This information improves the predictive power of machine learning models by extending the SOFA to include information related to gastrointestinal organ system. The described method improves mortality prediction by considering the dynamics of the extended SOFA score. Although tested on a limited data set, the results' stability across different models suggests robustness in real-time use. [ABSTRACT FROM AUTHOR]

The blood flow through the major vessels holds great diagnostic potential for the identification of cardiovascular complications and is therefore routinely assessed with current diagnostic modalities. Heart valves are subject to high hydrodynamic loads which render them prone to premature degradation. Failing native aortic valves are routinely replaced with bioprosthetic heart valves. This type of prosthesis is limited by a durability that is often less than the patient’s life expectancy. Frequent assessment of valvular function can therefore help to ensure good long-term outcomes and to plan reinterventions. In this article, we describe how unsupervised novelty detection algorithms can be used to automate the interpretation of blood flow data to improve outcomes through early detection of adverse cardiovascular events without requiring repeated check-ups in a clinical environment. The proposed method was tested in an in-vitro flow loop which allowed simulating a failing aortic valve in a laboratory setting. Aortic regurgitation of increasing severity was deliberately introduced with tube-shaped inserts, preventing complete valve closure during diastole. Blood flow recordings from a flow meter at the location of the ascending aorta were analyzed with the algorithms introduced in this article and a diagnostic index was defined that reflects the severity of valvular degradation. The results indicate that the proposed methodology offers a high sensitivity towards pathological changes of valvular function and that it is capable of automatically identifying valvular degradation. Such methods may be a step towards computer-assisted diagnostics and telemedicine that provide the clinician with novel tools to improve patient care. [ABSTRACT FROM AUTHOR]

Objective: The Centers for Disease Control and Prevention (CDC) coordinates a labor-intensive process to measure the prevalence of autism spectrum disorder (ASD) among children in the United States. Random forests methods have shown promise in speeding up this process, but they lag behind human classification accuracy by about 5%. We explore whether more recently available document classification algorithms can close this gap. Materials and methods: Using data gathered from a single surveillance site, we applied 8 supervised learning algorithms to predict whether children meet the case definition for ASD based solely on the words in their evaluations. We compared the algorithms’ performance across 10 random train-test splits of the data, using classification accuracy, F1 score, and number of positive calls to evaluate their potential use for surveillance. Results: Across the 10 train-test cycles, the random forest and support vector machine with Naive Bayes features (NB-SVM) each achieved slightly more than 87% mean accuracy. The NB-SVM produced significantly more false negatives than false positives (P = 0.027), but the random forest did not, making its prevalence estimates very close to the true prevalence in the data. The best-performing neural network performed similarly to the random forest on both measures. Discussion: The random forest performed as well as more recently available models like the NB-SVM and the neural network, and it also produced good prevalence estimates. NB-SVM may not be a good candidate for use in a fully-automated surveillance workflow due to increased false negatives. More sophisticated algorithms, like hierarchical convolutional neural networks, may not be feasible to train due to characteristics of the data. Current algorithms might perform better if the data are abstracted and processed differently and if they take into account information about the children in addition to their evaluations. Conclusion: Deep learning models performed similarly to traditional machine learning methods at predicting the clinician-assigned case status for CDC’s autism surveillance system. While deep learning methods had limited benefit in this task, they may have applications in other surveillance systems. [ABSTRACT FROM AUTHOR]

Determination and classification of the bruise degree for cherry can improve consumer satisfaction with cherry quality and enhance the industry’s competiveness and profitability. In this study, visible and near infrared (Vis-NIR) reflection spectroscopy was used for identifying bruise degree of cherry in 350–2500 nm. Sampling spectral data were extracted from normal, slight and severe bruise samples. Principal component analysis (PCA) was implemented to determine the first few principal components (PCs) for cluster analysis among samples. Optimal wavelengths were selected by loadings of PCs from PCA and successive projection algorithm (SPA) method, respectively. Afterwards, these optimal wavelengths were empolyed to establish the classification models as inputs of least square-support vector machine (LS-SVM). Better performance for qualitative discrimination of the bruise degree for cherry was emerged in LS-SVM model based on five optimal wavelengths (603, 633, 679, 1083, and 1803 nm) selected directly by SPA, which showed acceptable results with the classification accuracy of 93.3%. Confusion matrix illustrated misclassification generally occurred in normal and slight bruise samples. Furthermore, the latent relation between spectral property of cherries in varying bruise degree and its firmness and soluble solids content (SSC) was analyzed. The result showed both colour, firmness and SSC were consistent with the Vis-NIR reflectance of cherries. Overall, this study revealed that Vis-NIR reflection spectroscopy integrated with multivariate analysis can be used as a rapid, intact method to determine the bruise degree of cherry, laying a foundation for cherry sorting and postharvest quality control. [ABSTRACT FROM AUTHOR]

While most of the existing class stability assessors just rely on structural information retrieved from a desired source code snapshot. However, class stability is intrinsically characterized by the evolution of a number of dependencies and change propagation factors which aid to promote the ripple effect. Identification of classes prone to ripple effect (instable classes) through mining the version history of change propagation factors can aid developers to reduce the efforts needed to maintain and evolve the system. We propose Historical Information for Class Stability Prediction (HICSP), an approach to exploit change history information to predict the instable classes based on its correlation with change propagation factors. Subsequently, we performed two empirical studies. In the first study, we evaluate the HICSP on the version history of 10 open source projects. Subsequently, in the second replicated study, we evaluate the effectiveness of HICSP by tuning the parameters of its stability assessors. We observed the 4 to 16 percent improvement in term of F-measure value to predict the instable classes through HICSP as compared to existing class stability assessors. The promising results indicate that HICSP is able to identify instable classes and can aid developers in their decision making. [ABSTRACT FROM AUTHOR]

Short-term metro passenger flow forecasting is an essential component of intelligent transportation systems (ITS) and can be applied to optimize the passenger flow organization of a station and offer data support for metro passenger flow early warning and system management. LSTM neural networks have recently achieved remarkable recent in the field of natural language processing (NLP) because they are well suited for learning from experience to predict time series. For this purpose, we propose an empirical mode decomposition (EMD)-based long short-term memory (LSTM) neural network model for predicting short-term metro inbound passenger flow. The EMD algorithm decomposes the original sequential passenger flow into several intrinsic mode functions (IMFs) and a residual. Selected IMFs that are strongly correlated with the original data can be obtained via feature selection. The selected IMFs and the original data are integrated into inputs for LSTM neural networks, and a single LSTM prediction model and an EMD-LSTM hybrid forecasting model are developed. Finally, historical real automatic fare collection (AFC) data from metro passengers are collected from Chengdu Metro to verify the validity of the proposed EMD-LSTM prediction model. The results indicate that the proposed EMD-LSTM hybrid forecasting model outperforms the LSTM, ARIMA and BPN models. [ABSTRACT FROM AUTHOR]

Preeclampsia is one of the leading causes of maternal and fetal morbidity and mortality. Due to the lack of effective preventive measures, its prediction is essential to its prompt management. This study aimed to develop models using machine learning to predict late-onset preeclampsia using hospital electronic medical record data. The performance of the machine learning based models and models using conventional statistical methods were also compared. A total of 11,006 pregnant women who received antenatal care at Yonsei University Hospital were included. Maternal data were retrieved from electronic medical records during the early second trimester to 34 weeks. The prediction outcome was late-onset preeclampsia occurrence after 34 weeks’ gestation. Pattern recognition and cluster analysis were used to select the parameters included in the prediction models. Logistic regression, decision tree model, naïve Bayes classification, support vector machine, random forest algorithm, and stochastic gradient boosting method were used to construct the prediction models. C-statistics was used to assess the performance of each model. The overall preeclampsia development rate was 4.7% (474 patients). Systolic blood pressure, serum blood urea nitrogen and creatinine levels, platelet counts, serum potassium level, white blood cell count, serum calcium level, and urinary protein were the most influential variables included in the prediction models. C-statistics for the decision tree model, naïve Bayes classification, support vector machine, random forest algorithm, stochastic gradient boosting method, and logistic regression models were 0.857, 0.776, 0.573, 0.894, 0.924, and 0.806, respectively. The stochastic gradient boosting model had the best prediction performance with an accuracy and false positive rate of 0.973 and 0.009, respectively. The combined use of maternal factors and common antenatal laboratory data of the early second trimester through early third trimester could effectively predict late-onset preeclampsia using machine learning algorithms. Future prospective studies are needed to verify the clinical applicability algorithms. [ABSTRACT FROM AUTHOR]

As online content continues to grow, so does the spread of hate speech. We identify and examine challenges faced by online automatic approaches for hate speech detection in text. Among these difficulties are subtleties in language, differing definitions on what constitutes hate speech, and limitations of data availability for training and testing of these systems. Furthermore, many recent approaches suffer from an interpretability problem—that is, it can be difficult to understand why the systems make the decisions that they do. We propose a multi-view SVM approach that achieves near state-of-the-art performance, while being simpler and producing more easily interpretable decisions than neural methods. We also discuss both technical and practical challenges that remain for this task. [ABSTRACT FROM AUTHOR]

As limited resources, soils are the largest terrestrial sinks of organic carbon. In this respect, 3D modelling of soil organic carbon (SOC) offers substantial improvements in the understanding and assessment of the spatial distribution of SOC stocks. Previous three-dimensional SOC modelling approaches usually averaged each depth increment for multi-layer two-dimensional predictions. Therefore, these models are limited in their vertical resolution and thus in the interpretability of the soil as a volume as well as in the accuracy of the SOC stock predictions. So far, only few approaches used spatially modelled depth functions for SOC predictions. This study implemented and evaluated an approach that compared polynomial, logarithmic and exponential depth functions using non-linear machine learning techniques, i.e. multivariate adaptive regression splines, random forests and support vector machines to quantify SOC stocks spatially and depth-related in the context of biodiversity and ecosystem functioning research. The legacy datasets used for modelling include profile data for SOC and bulk density (BD), sampled at five depth increments (0-5, 5-10, 10-20, 20-30, 30-50 cm). The samples were taken in an experimental forest in the Chinese subtropics as part of the biodiversity and ecosystem functioning (BEF) China experiment. Here we compared the depth functions by means of the results of the different machine learning approaches obtained based on multi-layer 2D models as well as 3D models. The main findings were (i) that 3rd degree polynomials provided the best results for SOC and BD (R2 = 0.99 and R2 = 0.98; RMSE = 0.36% and 0.07 g cm-3). However, they did not adequately describe the general asymptotic trend of SOC and BD. In this respect the exponential (SOC: R2 = 0.94; RMSE = 0.56%) and logarithmic (BD: R2 = 84; RMSE = 0.21 g cm-3) functions provided more reliable estimates. (ii) random forests with the exponential function for SOC correlated better with the corresponding 2.5D predictions (R2: 0.96 to 0.75), compared to the 3rd degree polynomials (R2: 0.89 to 0.15) which support vector machines fitted best. We recommend not to use polynomial functions with sparsely sampled profiles, as they have many turning points and tend to overfit the data on a given profile. This may limit the spatial prediction capacities. Instead, less adaptive functions with a higher degree of generalisation such as exponential and logarithmic functions should be used to spatially map sparse vertical soil profile datasets. We conclude that spatial prediction of SOC using exponential depth functions, in conjunction with random forests is well suited for 3D SOC stock modelling, and provides much finer vertical resolutions compared to 2.5D approaches. [ABSTRACT FROM AUTHOR]

The thermoanalytical technique differential scanning calorimetry (DSC) has been applied to characterize protein denaturation patterns (thermograms) in blood plasma samples and relate these to a subject's health status. The analysis and classification of thermograms is challenging because of the high-dimensionality of the dataset. There are various methods for group classification using high-dimensional data sets; however, the impact of using high-dimensional data sets for cancer classification has been poorly understood. In the present article, we proposed a statistical approach for data reduction and a parametric method (PM) for modeling of high-dimensional data sets for two- and three- group classification using DSC and demographic data. We compared the PM to the non-parametric classification method K-nearest neighbors (KNN) and the semi-parametric classification method KNN with dynamic time warping (DTW). We evaluated the performance of these methods for multiple two-group classifications: (i) normal versus cervical cancer, (ii) normal versus lung cancer, (iii) normal versus cancer (cervical + lung), (iv) lung cancer versus cervical cancer as well as for three-group classification: normal versus cervical cancer versus lung cancer. In general, performance for two-group classification was high whereas three-group classification was more challenging, with all three methods predicting normal samples more accurately than cancer samples. Moreover, specificity of the PM method was mostly higher or the same as KNN and DTW-KNN with lower sensitivity. The performance of KNN and DTW-KNN decreased with the inclusion of demographic data, whereas similar performance was observed for the PM which could be explained by the fact that the PM uses fewer parameters as compared to KNN and DTW-KNN methods and is thus less susceptible to the risk of overfitting. More importantly the accuracy of the PM can be increased by using a greater number of quantile data points and by the inclusion of additional demographic and clinical data, providing a substantial advantage over KNN and DTW-KNN methods. [ABSTRACT FROM AUTHOR]

Introduction: Ovarian tumors are the most common diagnostic challenge for gynecologists and ultrasound examination has become the main technique for assessment of ovarian pathology and for preoperative distinction between malignant and benign ovarian tumors. However, ultrasonography is highly examiner-dependent and there may be an important variability between two different specialists when examining the same case. The objective of this work is the evaluation of different well-known Machine Learning (ML) systems to perform the automatic categorization of ovarian tumors from ultrasound images. Methods: We have used a real patient database whose input features have been extracted from 348 images, from the IOTA tumor images database, holding together with the class labels of the images. For each patient case and ultrasound image, its input features have been previously extracted using Fourier descriptors computed on the Region Of Interest (ROI). Then, four ML techniques are considered for performing the classification stage: K-Nearest Neighbors (KNN), Linear Discriminant (LD), Support Vector Machine (SVM) and Extreme Learning Machine (ELM). Results: According to our obtained results, the KNN classifier provides inaccurate predictions (less than 60% of accuracy) independently of the size of the local approximation, whereas the classifiers based on LD, SVM and ELM are robust in this biomedical classification (more than 85% of accuracy). Conclusions: ML methods can be efficiently used for developing the classification stage in computer-aided diagnosis systems of ovarian tumor from ultrasound images. These approaches are able to provide automatic classification with a high rate of accuracy. Future work should aim at enhancing the classifier design using ensemble techniques. Another ongoing work is to exploit different kind of features extracted from ultrasound images. [ABSTRACT FROM AUTHOR]

Demand for spring onion seeds is variable and maintaining its supply is crucial to the success of seed companies. Spring onion seed demand forecasting, which can help reduce the high operational costs increased by long-period propagation and complex logistics, has not previously been investigated yet. This paper provides a novel perspective on spring onion seed demand forecasting and proposes a hybrid Holt-Winters and support vector machine (SVM) forecasting model. The model uses dynamic factors, including historical seed sales, seed inventory, spring onion crop market price and weather data, as inputs to forecast spring onion seed demand. Forecasting error, i.e. the difference between actual and forecasted demand, is assessed. Two advanced machine learning models are trained on the same dataset as benchmark models. Numerical experiments using actual commercial sales data for three spring onion seed varieties show the proposed hybrid model outperformed the statistical-based models for all three forecasting errors. Seed inventory, spring onion crop market price and historical seed sales are the most important dynamic factors, among which seed inventory has short-term influence while other two have mid-term influence on seed demand forecasting. The absolute minimum temperature is the only factor having long-term influence. This study provides a promising spring onion seed demand forecasting model that helps understand the relationships between seed demand and other dynamic factors and the model could potentially be applied to demand forecasting of other crop seeds to reduce total operational costs. [ABSTRACT FROM AUTHOR]

In this paper, we propose a human action recognition method using HOIRM (histogram of oriented interest region motion) feature fusion and a BOW (bag of words) model based on AP (affinity propagation) clustering. First, a HOIRM feature extraction method based on spatiotemporal interest points ROI is proposed. HOIRM can be regarded as a middle-level feature between local and global features. Then, HOIRM is fused with 3D HOG and 3D HOF local features using a cumulative histogram. The method further improves the robustness of local features to camera view angle and distance variations in complex scenes, which in turn improves the correct rate of action recognition. Finally, a BOW model based on AP clustering is proposed and applied to action classification. It obtains the appropriate visual dictionary capacity and achieves better clustering effect for the joint description of a variety of features. The experimental results demonstrate that by using the fused features with the proposed BOW model, the average recognition rate is 95.75% in the KTH database, and 88.25% in the UCF database, which are both higher than those by using only 3D HOG+3D HOF or HOIRM features. Moreover, the average recognition rate achieved by the proposed method in the two databases is higher than that obtained by other methods. [ABSTRACT FROM AUTHOR]

Protein-protein interaction network data provides valuable information that infers direct links between genes and their biological roles. This information brings a fundamental hypothesis for protein function prediction that interacting proteins tend to have similar functions. With the help of recently-developed network embedding feature generation methods and deep maxout neural networks, it is possible to extract functional representations that encode direct links between protein-protein interactions information and protein function. Our novel method, STRING2GO, successfully adopts deep maxout neural networks to learn functional representations simultaneously encoding both protein-protein interactions and functional predictive information. The experimental results show that STRING2GO outperforms other protein-protein interaction network-based prediction methods and one benchmark method adopted in a recent large scale protein function prediction competition. [ABSTRACT FROM AUTHOR]

This paper attempts to design an automated, efficient and intelligent tomato grading method that facilitates the graded selling of the fruit. Based on machine vision, the color images of tomatoes with different morphologies were studied, and the color, shape and size were selected as the key features. On this basis, an automated grading classifier was created based on the surface features of tomatoes, and a grading platform was set up to verify the effect of the classifier. Specifically, the Hue value distributions of tomatoes with different maturities were investigated, and the Hue value ranges were determined for mature, semi-mature and immature tomatoes, producing the color classifier. Next, the first-order Fourier descriptor (1D- FD) was adopted to describe the radius sequence of tomato contour, and an equation was established to compute the irregularity of tomato contour, creating the shape classifier. After that, a linear regression equation was constructed to reflect the relationship between the transverse diameters of actual tomatoes and tomato images, and a classifier between large, medium and small tomatoes was produced based on the transverse diameter. Finally, a comprehensive tomato classifier was built based on the color, shape and size diameters. The experimental results show that the mean grading accuracy of the proposed method was 90.7%. This means our method can achieve automated real-time grading of tomatoes. [ABSTRACT FROM AUTHOR]

A key goal of precision medicine is predicting the best drug therapy for a specific patient from genomic information. In oncology, cancers that appear similar pathologically can vary greatly in how they respond to the same drug. Fortunately, data from high-throughput screening programs often reveal important relationships between genomic variability of cancer cells and their response to drugs. Nevertheless, many current computational methods to predict compound activity against cancer cells require large quantities of genomic, epigenomic, and additional cellular data to develop and to apply. Here we integrate recent screening data and machine learning to train classification models that predict the activity/inactivity of compounds against cancer cells based on the mutational status of only 145 oncogenes and a set of compound structural descriptors. Using IC50 values of 1 μM as activity cutoffs, our predictive models have sensitivities of 87%, specificities of 87%, and yield an area under the receiver operating characteristic curve equal to 0.94. We also develop regression models to predict log(IC50) values of compounds for cancer cells; the models achieve a Pearson correlation coefficient of 0.86 for cross-validation and up to 0.65–0.73 against blind test sets. Predictive performance remains strong when as few as 50 oncogenes are included. Finally, even when 40% of experimental IC50 values are missing from screening data, they can be imputed with sufficient reliability that classification accuracy is not diminished. The presented models are fast to generate and may serve as easily implemented screening tools for personalized oncology medicine, drug repurposing, and drug discovery. [ABSTRACT FROM AUTHOR]

The diagnosis and prognosis of patients with severe chronic disorders of consciousness are still challenging issues and a high rate of misdiagnosis is evident. Hence, new tools are needed for an accurate diagnosis, which will also have an impact on the prognosis. In recent years, functional Magnetic Resonance Imaging (fMRI) has been gaining more and more importance when diagnosing this patient group. Especially resting state scans, i.e., an examination when the patient does not perform any task in particular, seems to be promising for these patient groups. After preprocessing the resting state fMRI data with a standard pipeline, we extracted the correlation matrices of 132 regions of interest. The aim was to find the regions of interest which contributed most to the distinction between the different patient groups and healthy controls. We performed feature selection using a genetic algorithm and a support vector machine. Moreover, we show by using only those regions of interest for classification that are most often selected by our algorithm, we get a much better performance of the classifier. [ABSTRACT FROM AUTHOR]

Despite the rapid increase in the number and applications of plankton imaging systems in marine science, processing large numbers of images remains a major challenge due to large variations in image content and quality in different marine environments. We constructed an automatic plankton image recognition and enumeration system using an enhanced Convolutional Neural Network (CNN) and examined the performance of different network structures on automatic plankton image classification. The procedure started with an adaptive thresholding approach to extract Region of Interest (ROIs) from in situ plankton images, followed by a procedure to suppress the background noise and enhance target features for each extracted ROI. The enhanced ROIs were classified into seven categories by a pre-trained classifier which was a combination of a CNN and a Support Vector Machine (SVM). The CNN was selected to improve feature description and the SVM was utilized to improve classification accuracy. A series of comparison experiments were then conducted to test the effectiveness of the pre-trained classifier including the combination of CNN and SVM versus CNN alone, and the performance of different CNN models. Compared to CNN model alone, the combination of CNN and SVM increased classification accuracy and recall rate by 7.13% and 6.41%, respectively. Among the selected CNN models, the ResNet50 performed the best with accuracy and recall at 94.52% and 94.13% respectively. The present study demonstrates that deep learning technique can improve plankton image recognition and that the results can provide useful information on the selection of different CNN models for plankton recognition. The proposed algorithm could be generally applied to images acquired from different imaging systems. [ABSTRACT FROM AUTHOR]

Cellular frustrated models have been developed to describe how the adaptive immune system works. They are composed by independent agents that continuously pair and unpair depending on the information that one sub-set of these agents display. The emergent dynamics is sensitive to changes in the displayed information and can be used to detect anomalies, which can be important to accomplish the immune system main function of protecting the host. Therefore, it has been hypothesized that these models could be adequate to model the immune system activation. Likewise it has been hypothesized that these models could provide inspiration to develop new artificial intelligence algorithms for data mining applications. However, computational algorithms do not need to follow strictly the immunological reality. Here, we investigate efficient implementation strategies of these immune inspired ideas for anomaly detection applications and use real data to compare the performance of cellular frustration algorithms with standard implementations of one-class support vector machines and deep autoencoders. Our results demonstrate that more efficient implementations of cellular frustration algorithms are possible and also that cellular frustration algorithms can be advantageous for semi-supervised anomaly detection applications given their robustness and accuracy. [ABSTRACT FROM AUTHOR]

Non-invasive prenatal testing (NIPT) enables accurate detection of fetal chromosomal trisomies. The majority of publicly available computational methods for sequencing-based NIPT analyses rely on low-coverage whole-genome sequencing (WGS) data and are not applicable for targeted high-coverage sequencing data from cell-free DNA samples. Here, we present a novel computational framework for a targeted high-coverage sequencing-based NIPT analysis. The developed framework uses a hidden Markov model (HMM) in conjunction with a supplemental machine learning model, such as decision tree (DT) or support vector machine (SVM), to detect fetal trisomy and parental origin of additional fetal chromosomes. These models were developed using simulated datasets covering a wide range of biologically relevant scenarios with various chromosomal quantities, parental origins of extra chromosomes, fetal DNA fractions, and sequencing read depths. Developed models were tested on simulated and experimental targeted sequencing datasets. Consequently, we determined the functional feasibility and limitations of each proposed approach and demonstrated that read count-based HMM achieved the best overall classification accuracy of 0.89 for detecting fetal euploidies and trisomies on simulated dataset. Furthermore, we show that by using the DT and SVM on the HMM classification results, it was possible to increase the final trisomy classification accuracy to 0.98 and 0.99, respectively. We demonstrate that read count and allelic ratio-based models can achieve a high accuracy (up to 0.98) for detecting fetal trisomy even if the fetal fraction is as low as 2%. Currently, existing commercial NIPT analysis requires at least 4% of fetal fraction, which can be possibly a challenge in case of early gestational age (<10 weeks) or high maternal body mass index (>35 kg/m2). More accurate detection can be achieved at higher sequencing depth using HMM in conjunction with supplemental models, which significantly improve the trisomy detection especially in borderline scenarios (e.g., very low fetal fraction) and enables to perform NIPT even earlier than 10 weeks of pregnancy. [ABSTRACT FROM AUTHOR]

Optical coherence tomography (OCT) based measurements of retinal layer thickness, such as the retinal nerve fibre layer (RNFL) and the ganglion cell with inner plexiform layer (GCIPL) are commonly employed for the diagnosis and monitoring of glaucoma. Previously, machine learning techniques have relied on segmentation-based imaging features such as the peripapillary RNFL thickness and the cup-to-disc ratio. Here, we propose a deep learning technique that classifies eyes as healthy or glaucomatous directly from raw, unsegmented OCT volumes of the optic nerve head (ONH) using a 3D Convolutional Neural Network (CNN). We compared the accuracy of this technique with various feature-based machine learning algorithms and demonstrated the superiority of the proposed deep learning based method. Logistic regression was found to be the best performing classical machine learning technique with an AUC of 0.89. In direct comparison, the deep learning approach achieved a substantially higher AUC of 0.94 with the additional advantage of providing insight into which regions of an OCT volume are important for glaucoma detection. Computing Class Activation Maps (CAM), we found that the CNN identified neuroretinal rim and optic disc cupping as well as the lamina cribrosa (LC) and its surrounding areas as the regions significantly associated with the glaucoma classification. These regions anatomically correspond to the well established and commonly used clinical markers for glaucoma diagnosis such as increased cup volume, cup diameter, and neuroretinal rim thinning at the superior and inferior segments. [ABSTRACT FROM AUTHOR]

Endosomes are subcellular organelles which serve as important transport compartments in eukaryotic cells. Fluorescence microscopy is a widely applied technology to study endosomes at the subcellular level. In general, a microscopy image can contain a large number of organelles and endosomes in particular. Detecting and annotating endosomes in fluorescence microscopy images is a critical part in the study of subcellular trafficking processes. Such annotation is usually performed by human inspection, which is time-consuming and prone to inaccuracy if carried out by inexperienced analysts. This paper proposes a two-stage method for automated detection of ring-like endosomes. The method consists of a localization stage cascaded by an identification stage. Given a test microscopy image, the localization stage generates a voting-map by locally comparing the query endosome patches and the test image based on a bag-of-words model. Using the voting-map, a number of candidate patches of endosomes are determined. Subsequently, in the identification stage, a support vector machine (SVM) is trained using the endosome patches and the background pattern patches. Each of the candidate patches is classified by the SVM to rule out those patches of endosome-like background patterns. The performance of the proposed method is evaluated with real microscopy images of human myeloid endothelial cells. It is shown that the proposed method significantly outperforms several state-of-the-art competing methods using multiple performance metrics. [ABSTRACT FROM AUTHOR]

Short-term traffic speed prediction is a key component of proactive traffic control in the intelligent transportation systems. The objective of this study is to investigate the short-term traffic speed prediction under different data collection time intervals. Traffic speed data was collected from an urban freeway in Edmonton, Canada. A seasonal autoregressive integrated moving average plus seasonal discrete grey model structure (SARIMA-SDGM) was proposed to perform the traffic speed prediction. The model performance of SARIMA-SDGM model was compared with that of the seasonal autoregressive integrated moving average (SARIMA) model, seasonal discrete grey model (SDGM), artificial neural network (ANN) model, and support vector regression (SVR) model. The results showed that SARIMA-SDGM model performs best with the lowest mean absolute error (MAE), mean absolute percentage error (MAPE), and the root mean square error (RMSE). The traffic speed prediction accuracy under different time intervals were compared based on the SARIMA-SDGM model. The results showed that the prediction accuracy improves with the increase in time interval. In addition, when the time interval is greater than 10 min, the prediction results yield stable prediction accuracy. [ABSTRACT FROM AUTHOR]

The advent of smartphones has advanced the use of embedded sensors to acquire various physiological information. For example, smartphone camera sensors and accelerometers can provide heart rhythm signals to the subjects, while microphones can give respiratory signals. However, the acquired smartphone-based physiological signals are more vulnerable to motion and noise artifacts (MNAs) compared to using medical devices, since subjects need to hold the smartphone with proper contact to the smartphone camera and lens stably and tightly for a duration of time without any movement in the hand or finger. This results in more MNA than traditional methods, such as placing a finger inside a tightly enclosed pulse oximeter to get PPG signals, which provides stable contact between the sensor and the subject’s finger. Moreover, a smartphone lens does not block ambient light in an effective way, while pulse oximeters are designed to block the ambient light effectively. In this paper, we propose a novel diversity method for smartphone signals that reduces the effect of MNAs during heart rhythm signal detection by 1) acquiring two heterogeneous signals from a color intensity signal and a fingertip movement signal, and 2) selecting the less MNA-corrupted signal of the two signals. The proposed method has advantages in that 1) diversity gain can be obtained from the two heterogeneous signals when one signal is clean while the other signal is corrupted, and 2) acquisition of the two heterogeneous signals does not double the acquisition procedure but maintains a single acquisition procedure, since two heterogeneous signals can be obtained from a single smartphone camera recording. In our diversity method, we propose to choose the better signal based on the signal quality indices (SQIs), i.e., standard deviation of instantaneous heart rate (STD–HR), root mean square of the successive differences of peak-to-peak time intervals (RMSSD–T), and standard deviation of peak values (STD–PV). As a performance metric evaluating the proposed diversity method, the ratio of usable period is considered. Experimental results show that our diversity method increases the usable period 19.53% and 6.25% compared to the color intensity or the fingertip movement signals only, respectively. [ABSTRACT FROM AUTHOR]

A brain-computer interface (BCI), based on motor imagery EEG, uses information extracted from the electroencephalography signals generated by a person who intends to perform any action. One of the most important issues of current research is how to detect automatically whether the user intends to send some message to a certain device. This study presents a proposal, based on a hierarchical structured system, for recognising intentional and non-intentional mental tasks on a BCI system by applying machine learning techniques to the EEG signals. First-level clustering is performed to distinguish between intentional control (IC) and non-intentional control (NC) state patterns. Then, the patterns recognised as IC are passed on to a second stage where supervised learning techniques are used to classify them. In BCI applications, it is critical to correctly classify NC states with a low false positive rate (FPR) to avoid undesirable effects. According to the literature, we selected a maximum FPR of 10%. Under these conditions, our proposal achieved an average test accuracy of 66.6%, with an 8.2% FPR, for the BCI competition IIIa dataset. The main contribution of this paper is the hierarchical approach, based on machine learning paradigms, which performs intentional and non-intentional discrimination and, depending on the case, classifies the intended command selected by the user. [ABSTRACT FROM AUTHOR]

Non-negative matrix tri-factorization (NMTF) is a popular technique for learning low-dimensional feature representation of relational data. Currently, NMTF learns a representation of a dataset through an optimization procedure that typically uses multiplicative update rules. This procedure has had limited success, and its failure cases have not been well understood. We here perform an empirical study involving six large datasets comparing multiplicative update rules with three alternative optimization methods, including alternating least squares, projected gradients, and coordinate descent. We find that methods based on projected gradients and coordinate descent converge up to twenty-four times faster than multiplicative update rules. Furthermore, alternating least squares method can quickly train NMTF models on sparse datasets but often fails on dense datasets. Coordinate descent-based NMTF converges up to sixteen times faster compared to well-established methods. [ABSTRACT FROM AUTHOR]

Objective: Suicide in adolescents is a major problem worldwide and previous history of suicide ideation and attempt represents the strongest predictors of future suicidal behavior. The aim of this study was to develop prediction model to identify Korean adolescents of high risk suicide (= who have history of suicide ideation/attempt in previous year) using machine learning techniques. Methods: A nationally representative dataset of Korea Youth Risk Behavior Web-based Survey (KYRBWS) was used (n = 59,984 of middle and high school students in 2017). The classification process was performed using machine learning techniques such as logistic regression (LR), random forest (RF), support vector machine (SVM), artificial neural network (ANN), and extreme gradient boosting (XGB). Results: A total of 7,443 adolescents (12.4%) had a previous history of suicidal ideation/attempt. In the multivariable analysis, sadness (odds ratio [OR], 6.41; 95% confidence interval [95% CI], 6.08–6.87), violence (OR, 2.32; 95% CI, 2.01–2.67), substance use (OR, 1.93; 95% CI, 1.52–2.45), and stress (OR, 1.63; 95% CI, 1.40–1.86) were associated factors. Taking into account 26 variables as predictors, the accuracy of models of machine learning techniques to predict the high-risk suicidal was comparable with that of LR; the accuracy was best in XGB (79.0%), followed by SVM (78.7%), LR (77.9%), RF (77.8%), and ANN (77.5%). Conclusions: The machine leaning techniques showed comparable performance with LR to classify adolescents who have previous history of suicidal ideation/attempt. This model will hopefully serve as a foundation for decreasing future suicides as it enables early identification of adolescents at risk of suicide and modification of risk factors. [ABSTRACT FROM AUTHOR]

Mangrove forests, one of the world’s most endangered ecosystems, are also some of the most difficult to access. This is especially true along the Pacific coast of Costa Rica, where 99% of the country’s mangroves occur. Unmanned Aerial Systems (UAS), or drones, have become a convenient tool for natural area assessment, and offer a solution to the problems of remote mangrove monitoring. This study is the first to use UAS to analyze the structure of a mangrove forests within Central America. Our goals were to (1) determine the forest structure of two estuaries in northwestern Costa Rica through traditional ground measurements, (2) assess the accuracy of UAS measurements of canopy height and percent coverage and (3) determine whether the normalized difference vegetation index (NDVI) could discriminate between the most abundant mangrove species. We flew a UAS equipped with a single NDVI sensor during the peak wet (Sept–Nov) and dry (Jan–Feb) seasons. The structure and species composition of the estuaries showed a possible transition between the wet mangroves of southern Costa Rica and the drier northern mangroves. UAS-derived measurements at 100 cm/pixel resolution of percent canopy coverage and maximum and mean canopy height were not statistically different from ground measurements (p > 0.05). However, there were differences in mean canopy height at 10 cm/pixel resolution (p = 0.043), indicating diminished returns in accuracy as resolution becomes extremely fine. Mean NDVI values of Avicennia germinans (most abundant species) changed significantly between seasons (p < 0.001). Mean NDVI of Rhizophora racemosa (second most abundant species) was significantly different from A. germinans and dry forest dominant plots during the dry season (p < 0.001), demonstrating NDVI’s capability of discriminating mangrove species. This study provides the first structural assessment of the studied estuaries and a framework for future studies of mangroves using UAS. [ABSTRACT FROM AUTHOR]

Owing to the emergence of the Internet and its rapid growth, people can use mobile devices on many social media platforms (blogs, Facebook forums, etc.), and the platforms provide well-known websites for people to express and share their daily activities and ideas on global issues. Many consumers utilize product review websites before making a purchase. Many well-known websites are searched for relevant product reviews and experiences of product use. We can easily collect large amounts of structured and unstructured product data and further analyze the data to determine the desired product information. For this reason, many researchers are gradually focusing on sentiment analysis or opinion exploration (opinion mining) and use this technique to extract and analyze customer opinions and emotions. This paper proposes a sentimental text mining method based on an additional features method to enhance accuracy and reduce implementation time and uses singular value decomposition and principal component analysis for data dimension reduction. This study has four contributions: (1) the proposed algorithm for preprocessing the data for sentiment classification, (2) the additional features to enhance the accuracy of the sentiment classification, (3) the application of singular value decomposition and principal component analysis for data dimension reduction, and (4) the design of five modules based on different features, with or without stemming, to compare the performance results. The experimental results show that the proposed method has better accuracy than other methods and that the proposed method can decrease the implementation time. [ABSTRACT FROM AUTHOR]

Accurately predicting reference evapotranspiration (ET0) with limited climatic data is crucial for irrigation scheduling design and agricultural water management. This study evaluated eight machine learning models in four categories, i.e. neuron-based (MLP, GRNN and ANFIS), kernel-based (SVM, KNEA), tree-based (M5Tree, XGBoost) and curve-based (MARS) models, for predicting daily ET0 with maximum/maximum temperature and precipitation data during 2001–2015 from 14 stations in various climatic regions of China, i.e., arid desert of northwest China (NWC), semi-arid steppe of Inner Mongolia (IM), Qinghai-Tibetan Plateau (QTP), (semi-)humid cold-temperate northeast China (NEC), semi-humid warm-temperate north China (NC), humid subtropical central China (CC) and humid tropical south China (SC). The results showed machine learning models using only temperature data obtained satisfactory daily ET0 estimates (on average R2 = 0.829, RMSE = 0.718 mm day−1, NRMSE = 0.250 and MAE = 0.508 mm day−1). The prediction accuracy was improved by 7.6% across China when information of precipitation was further considered, particularly in (sub)tropical humid regions (by 9.7% in CC and 12.4% in SC). The kernel-based SVM, KNEA and curve-based MARS models generally outperformed the others in terms of prediction accuracy, with the best performance by KNEA in NWC and IM, by SVM in QTP, CC and SC, and very similar performance by them in NEC and NC. SVM (1.9%), MLP (2.0%), MARS (2.6%) and KNEA (6.4%) showed relatively small average increases in RMSE during testing compared with training RMSE. SVM is highly recommended for predicting daily ET0 across China in light of best accuracy and stability, while KNEA and MARS are also promising powerful models. [ABSTRACT FROM AUTHOR]

Reference evapotranspiration (ET0) plays a fundamental role in irrigated agriculture. The objective of this study is to simulate monthly ET0 at a meteorological station in India using a new method, an improved support vector machine (SVM) based on the cuckoo algorithm (CA), which is known as SVM-CA. Maximum temperature, minimum temperature, relative humidity, wind speed and sunshine hours were selected as inputs for the models used in the simulation. The results of the simulation using SVM-CA were compared with those from experimental models, genetic programming (GP), model tree (M5T) and the adaptive neuro-fuzzy inference system (ANFIS). The achieved results demonstrate that the proposed SVM-CA model is able to simulate ET0 more accurately than the GP, M5T and ANFIS models. Two major indicators, namely, root mean square error (RMSE) and mean absolute error (MAE), indicated that the SVM-CA outperformed the other methods with respective reductions of 5–15% and 5–17% compared with the GP model, 12–21% and 10–22% compared with the M5T model, and 7–15% and 5–18% compared with the ANFIS model, respectively. Therefore, the proposed SVM-CA model has high potential for accurate simulation of monthly ET0 values compared with the other models. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy is a complication of diabetes that produces changes in the blood vessel structure in the retina, which can cause severe vision problems and even blindness. In this paper, we demonstrate that by identifying topological features in very high resolution retinal images, we can construct a classifier that discriminates between healthy patients and those with diabetic retinopathy using summary statistics of these features. Topological data analysis identifies the features as connected components and holes in the images and describes the extent to which they persist across the image. These features are encoded in persistence diagrams, summaries of which can be used to discrimate between diabetic and healthy patients. The method has the potential to be an effective automated screening tool, with high sensitivity and specificity. [ABSTRACT FROM AUTHOR]

Early defibrillation by an automated external defibrillator (AED) is key for the survival of out-of-hospital cardiac arrest (OHCA) patients. ECG feature extraction and machine learning have been successfully used to detect ventricular fibrillation (VF) in AED shock decision algorithms. Recently, deep learning architectures based on 1D Convolutional Neural Networks (CNN) have been proposed for this task. This study introduces a deep learning architecture based on 1D-CNN layers and a Long Short-Term Memory (LSTM) network for the detection of VF. Two datasets were used, one from public repositories of Holter recordings captured at the onset of the arrhythmia, and a second from OHCA patients obtained minutes after the onset of the arrest. Data was partitioned patient-wise into training (80%) to design the classifiers, and test (20%) to report the results. The proposed architecture was compared to 1D-CNN only deep learners, and to a classical approach based on VF-detection features and a support vector machine (SVM) classifier. The algorithms were evaluated in terms of balanced accuracy (BAC), the unweighted mean of the sensitivity (Se) and specificity (Sp). The BAC, Se, and Sp of the architecture for 4-s ECG segments was 99.3%, 99.7%, and 98.9% for the public data, and 98.0%, 99.2%, and 96.7% for OHCA data. The proposed architecture outperformed all other classifiers by at least 0.3-points in BAC in the public data, and by 2.2-points in the OHCA data. The architecture met the 95% Sp and 90% Se requirements of the American Heart Association in both datasets for segment lengths as short as 3-s. This is, to the best of our knowledge, the most accurate VF detection algorithm to date, especially on OHCA data, and it would enable an accurate shock no shock diagnosis in a very short time. [ABSTRACT FROM AUTHOR]

The Text Forum Threads (TFThs) contain a large amount of Initial-Posts Replies pairs (IPR pairs) which are related to information exchange and discussion amongst the forum users with similar interests. Generally, some user replies in the discussion thread are off-topic and irrelevant. Hence, the content is of different qualities. It is important to identify the quality of the IPR pairs in a discussion thread in order to extract relevant information and helpful replies because a higher frequency of irrelevant replies in the thread could take the discussion in a different direction and the genuine users would lose interest in this discussion thread. In this study, the authors have presented an approach for identifying the high-quality user replies to the Initial-Post and use some quality dimensions features for their extraction. Moreover, crowdsourcing platforms were used for judging the quality of the replies and classified them into high-quality, low-quality or non-quality replies to the Initial-Posts. Then, the high-quality IPR pairs were extracted and identified based on their quality, and they were ranked using three classifiers i.e., Support Vector Machine, Naïve Bayes, and the Decision Trees according to their quality dimensions of relevancy, author activeness, timeliness, ease-of-understanding, politeness, and amount-of-data. In conclusion, the experimental results for the TFThs showed that the proposed approach could improve the extraction of the quality replies and identify the quality features that can be used for the Text Forum Thread Summarization. [ABSTRACT FROM AUTHOR]

Background: Identifying people at risk of cardiovascular diseases (CVD) is a cornerstone of preventative cardiology. Risk prediction models currently recommended by clinical guidelines are typically based on a limited number of predictors with sub-optimal performance across all patient groups. Data-driven techniques based on machine learning (ML) might improve the performance of risk predictions by agnostically discovering novel risk predictors and learning the complex interactions between them. We tested (1) whether ML techniques based on a state-of-the-art automated ML framework (AutoPrognosis) could improve CVD risk prediction compared to traditional approaches, and (2) whether considering non-traditional variables could increase the accuracy of CVD risk predictions. Methods and findings: Using data on 423,604 participants without CVD at baseline in UK Biobank, we developed a ML-based model for predicting CVD risk based on 473 available variables. Our ML-based model was derived using AutoPrognosis, an algorithmic tool that automatically selects and tunes ensembles of ML modeling pipelines (comprising data imputation, feature processing, classification and calibration algorithms). We compared our model with a well-established risk prediction algorithm based on conventional CVD risk factors (Framingham score), a Cox proportional hazards (PH) model based on familiar risk factors (i.e, age, gender, smoking status, systolic blood pressure, history of diabetes, reception of treatments for hypertension and body mass index), and a Cox PH model based on all of the 473 available variables. Predictive performances were assessed using area under the receiver operating characteristic curve (AUC-ROC). Overall, our AutoPrognosis model improved risk prediction (AUC-ROC: 0.774, 95% CI: 0.768-0.780) compared to Framingham score (AUC-ROC: 0.724, 95% CI: 0.720-0.728, p < 0.001), Cox PH model with conventional risk factors (AUC-ROC: 0.734, 95% CI: 0.729-0.739, p < 0.001), and Cox PH model with all UK Biobank variables (AUC-ROC: 0.758, 95% CI: 0.753-0.763, p < 0.001). Out of 4,801 CVD cases recorded within 5 years of baseline, AutoPrognosis was able to correctly predict 368 more cases compared to the Framingham score. Our AutoPrognosis model included predictors that are not usually considered in existing risk prediction models, such as the individuals’ usual walking pace and their self-reported overall health rating. Furthermore, our model improved risk prediction in potentially relevant sub-populations, such as in individuals with history of diabetes. We also highlight the relative benefits accrued from including more information into a predictive model (information gain) as compared to the benefits of using more complex models (modeling gain). Conclusions: Our AutoPrognosis model improves the accuracy of CVD risk prediction in the UK Biobank population. This approach performs well in traditionally poorly served patient subgroups. Additionally, AutoPrognosis uncovered novel predictors for CVD disease that may now be tested in prospective studies. We found that the “information gain” achieved by considering more risk factors in the predictive model was significantly higher than the “modeling gain” achieved by adopting complex predictive models. [ABSTRACT FROM AUTHOR]

Objective: To compare axonal loss in ganglion cells detected with swept-source optical coherence tomography (SS-OCT) in eyes of patients with multiple sclerosis (MS) versus healthy controls using different machine learning techniques. To analyze the capability of machine learning techniques to improve the detection of retinal nerve fiber layer (RNFL) and the complex Ganglion Cell Layer–Inner plexiform layer (GCL+) damage in patients with multiple sclerosis and to use the SS-OCT as a biomarker to early predict this disease. Methods: Patients with relapsing-remitting MS (n = 80) and age-matched healthy controls (n = 180) were enrolled. Different protocols from the DRI SS-OCT Triton system were used to obtain the RNFL and GCL+ thicknesses in both eyes. Macular and peripapilar areas were analyzed to detect the zones with higher thickness decrease. The performance of different machine learning techniques (decision trees, multilayer perceptron and support vector machine) for identifying RNFL and GCL+ thickness loss in patients with MS were evaluated. Receiver-operating characteristic (ROC) curves were used to display the ability of the different tests to discriminate between MS and healthy eyes in our population. Results: Machine learning techniques provided an excellent tool to predict MS disease using SS-OCT data. In particular, the decision trees obtained the best prediction (97.24%) using RNFL data in macular area and the area under the ROC curve was 0.995, while the wide protocol which covers an extended area between macula and papilla gave an accuracy of 95.3% with a ROC of 0.998. Moreover, it was obtained that the most significant area of the RNFL to predict MS is the macula just surrounding the fovea. On the other hand, in our study, GCL+ did not contribute to predict MS and the different machine learning techniques performed worse in this layer than in RNFL. Conclusions: Measurements of RNFL thickness obtained with SS-OCT have an excellent ability to differentiate between healthy controls and patients with MS. Thus, the use of machine learning techniques based on these measures can be a reliable tool to help in MS diagnosis. [ABSTRACT FROM AUTHOR]

Specific entity terms such as disease, test, symptom, and genes in Electronic Medical Record (EMR) can be extracted by Named Entity Recognition (NER). However, limited resources of labeled EMR pose a great challenge for mining medical entity terms. In this study, a novel multitask bi-directional RNN model combined with deep transfer learning is proposed as a potential solution of transferring knowledge and data augmentation to enhance NER performance with limited data. The proposed model has been evaluated using micro average F-score, macro average F-score and accuracy. It is observed that the proposed model outperforms the baseline model in the case of discharge datasets. For instance, for the case of discharge summary, the micro average F-score is improved by 2.55% and the overall accuracy is improved by 7.53%. For the case of progress notes, the micro average F-score and the overall accuracy are improved by 1.63% and 5.63%, respectively. [ABSTRACT FROM AUTHOR]

It remains unclear how the visual system is able to extract affective content from complex scenes even with extremely brief (< 100 millisecond) exposures. One possibility, suggested by findings in machine vision, is that low-level features such as unlocalized, two-dimensional (2-D) Fourier spectra can be diagnostic of scene content. To determine whether Fourier image amplitude carries any information about the affective quality of scenes, we first validated the existence of image category differences through a support vector machine (SVM) model that was able to discriminate our intact aversive and neutral images with ~ 70% accuracy using amplitude-only features as inputs. This model allowed us to confirm that scenes belonging to different affective categories could be mathematically distinguished on the basis of amplitude spectra alone. The next question is whether these same features are also exploited by the human visual system. Subsequently, we tested observers’ rapid classification of affective and neutral naturalistic scenes, presented briefly (~33.3 ms) and backward masked with synthetic textures. We tested categorization accuracy across three distinct experimental conditions, using: (i) original images, (ii) images having their amplitude spectra swapped within a single affective image category (e.g., an aversive image whose amplitude spectrum has been swapped with another aversive image) or (iii) images having their amplitude spectra swapped between affective categories (e.g., an aversive image containing the amplitude spectrum of a neutral image). Despite its discriminative potential, the human visual system does not seem to use Fourier amplitude differences as the chief strategy for affectively categorizing scenes at a glance. The contribution of image amplitude to affective categorization is largely dependent on interactions with the phase spectrum, although it is impossible to completely rule out a residual role for unlocalized 2-D amplitude measures. [ABSTRACT FROM AUTHOR]

Based on electrohysterogram, this paper designed a new method using wavelet-based nonlinear features and stacked sparse autoencoder for preterm birth detection. For each sample, three level wavelet decomposition of a time series was performed. Approximation coefficients at level 3 and detail coefficients at levels 1, 2 and 3 were extracted. Sample entropy of the detail coefficients at levels 1, 2, 3 and approximation coefficients at level 3 were computed as features. The classifier was constructed based on stacked sparse autoencoder. In addition, stacked sparse autoencoder was further compared with extreme learning machine and support vector machine in relation to their classification performance of electrohysterogram. The experiment results reveal that classifier based on stacked sparse autoencoder showed better performance than the other two classifiers with an accuracy of 90%, a sensitivity of 92%, a specificity of 88%. The results indicate that the method proposed in this paper could be effective for detecting preterm birth in electrohysterogram and the framework designed in this work presents higher discriminability than other techniques. [ABSTRACT FROM AUTHOR]

The fragile ecological environment near mines provide advantageous conditions for the development of landslides. Mine landslide susceptibility mapping is of great importance for mine geo-environment control and restoration planning. In this paper, a total of 493 landslides in Shangli County, China were collected through historical landslide inventory. 16 spectral, geomorphic and hydrological predictive factors, mainly derived from Landsat 8 imagery and Global Digital Elevation Model (ASTER GDEM), were prepared initially for landslide susceptibility assessment. Predictive capability of these factors was evaluated by using the value of variance inflation factor and information gain ratio. Three models, namely artificial neural network (ANN), support vector machine (SVM) and information value model (IVM), were applied to assess the mine landslide sensitivity. The receiver operating characteristic curve (ROC) and rank probability score were used to validate and compare the comprehensive predictive capabilities of three models involving uncertainty. Results showed that ANN model achieved higher prediction capability, proving its advantage of solve nonlinear and complex problems. Comparing the estimated landslide susceptibility map with the ground-truth one, the high-prone area tends to be located in the middle area with multiple fault distributions and the steeply sloped hill. [ABSTRACT FROM AUTHOR]

Multivariate analysis methods have been widely applied to decode brain states from functional magnetic resonance imaging (fMRI) data. Among various multivariate analysis methods, partial least squares regression (PLSR) is often used to select relevant features for decoding brain states. However, PLSR is seldom directly used as a classifier to decode brain states from fMRI data. It is unclear how PLSR classifiers perform in brain-state decoding using fMRI. In this study, we propose two types of two-step PLSR classifiers that use PLSR/sparse PLSR (SPLSR) to select features and PLSR for classification to improve the performance of the PLSR classifier. The results of simulated and real fMRI data demonstrated that the PLSR classifier using PLSR/SPLSR to select features outperformed both the PLSR classifier using a general linear model (GLM) and the support vector machine (SVM) using PLSR/SPLSR/GLM in most cases. Moreover, PLSR using SPLSR to select features showed the best performance among all of the methods. Compared to GLM, PLSR is more sensitive in selecting the voxels that are specific to each task. The results suggest that the performance of the PLSR classifier can be largely improved when the PLSR classifier is combined with the feature selection methods of SPLSR and PLSR. [ABSTRACT FROM AUTHOR]

An image classification algorithm based on adaptive feature weight updating is proposed to address the low classification accuracy of the current single-feature classification algorithms and simple multifeature fusion algorithms. The MapReduce parallel programming model on the Hadoop platform is used to perform an adaptive fusion of hue, local binary pattern (LBP) and scale-invariant feature transform (SIFT) features extracted from images to derive optimal combinations of weights. The support vector machine (SVM) classifier is then used to perform parallel training to obtain the optimal SVM classification model, which is then tested. The Pascal VOC 2012, Caltech 256 and SUN databases are adopted to build a massive image library. The speedup, classification accuracy and training time are tested in the experiment, and the results show that a linear growth tendency is present in the speedup of the system in a cluster environment. In consideration of the hardware costs, time, performance and accuracy, the algorithm is superior to mainstream classification algorithms, such as the power mean SVM and convolutional neural network (CNN). As the number and types of images both increase, the classification accuracy rate exceeds 95%. When the number of images reaches 80,000, the training time of the proposed algorithm is only 1/5 that of traditional single-node architecture algorithms. This result reflects the effectiveness of the algorithm, which provides a basis for the effective analysis and processing of image big data. [ABSTRACT FROM AUTHOR]

Conventional traffic crash analyzing methods focus on identifying the relationship between traffic crash outcomes and impact risk factors and explaining the effects of risk factors, which ignore the changes of roadway systems and can lead to inaccurate results in traffic crash predictions. To address this issue, an innovative two-step method is proposed and a support vector regression (SVR) model is formulated into state-space model (SSM) framework for traffic crash prediction. The SSM was developed in the first step to identify the dynamic evolution process of the roadway systems that are caused by the changes of traffic flow and predict the changes of impact factors in roadway systems. Using the predicted impact factors, the SVR model was incorporated in the second step to perform the traffic crash prediction. A five-year dataset that obtained from 1152 roadway segments in Tennessee was employed to validate the model effectiveness. The proposed models result in an average prediction MAPE of 7.59%, a MAE of 0.11, and a RMSD of 0.32. For the performance comparison, a SVR model and a multivariate negative binomial (MVNB) model were developed to do the same task. The results show that the proposed model has superior performances in terms of prediction accuracy compared to the SVR and MVNB models. Compared to the SVR and MVNB models, the benefit of incorporating a state-space model to identify the changes of roadway systems is significant evident in the proposed models for all crash types, and the prediction accuracy that measured by MAPE can be improved by 4.360% and 6.445% on average, respectively. Apart from accuracy improvement, the proposed models are more robust and the predictions can retain a smoother pattern. Furthermore, the results show that the proposed model has a more precise and synchronized response behavior to the high variations of the observed data, especially for the phenomenon of extra zeros. [ABSTRACT FROM AUTHOR]

Degenerative cervical myelopathy (DCM) is a spinal cord condition that results in progressive non-traumatic compression of the cervical spinal cord. Spine surgeons must consider a large quantity of information relating to disease presentation, imaging features, and patient characteristics to determine if a patient will benefit from surgery for DCM. We applied a supervised machine learning approach to develop a classification model to predict individual patient outcome after surgery for DCM. Patients undergoing surgery for DCM as a part of the AOSpine CSM-NA or CSM-I prospective, multi-centre studies were included in the analysis. Out of 757 patients 605, 583, and 539 patients had complete follow-up information at 6, 12, and 24 months respectively and were included in the analysis. The primary outcome was improvement in the SF-6D quality of life indicator score by the minimum clinically important difference (MCID). The secondary outcome was improvement in the modified Japanese Orthopedic Association (mJOA) score by the MCID. Predictor variables reflected information about pre-operative disease severity, disease presentation, patient demographics, and comorbidities. A machine learning approach of feature engineering, data pre-processing, and model optimization was used to create the most accurate predictive model of outcome after surgery for DCM. Following data pre-processing 48, 108, and 101 features were chosen for model training at 6, 12, and 24 months respectively. The best performing predictive model used a random forest structure and had an average area under the curve (AUC) of 0.70, classification accuracy of 77%, and sensitivity of 78% when evaluated on a testing cohort that was not used for model training. Worse pre-operative disease severity, longer duration of DCM symptoms, older age, higher body weight, and current smoking status were associated with worse surgical outcomes. We developed a model that predicted positive surgical outcome for DCM with good accuracy at the individual patient level on an independent testing cohort. Our analysis demonstrates the applicability of machine-learning to predictive modeling in spine surgery. [ABSTRACT FROM AUTHOR]

Depressive symptoms are related to abnormalities in the autonomic nervous system (ANS), and physiological signals that can be used to measure and evaluate such abnormalities have previously been used as indicators for diagnosing mental disorder, such as major depressive disorder (MDD). In this study, we investigate the feasibility of developing an objective measure of depressive symptoms that is based on examining physiological abnormalities in individuals when they are experiencing mental stress. To perform this, we recruited 30 patients with MDD and 31 healthy controls. Then, skin conductance (SC) was measured during five 5-min experimental phases, comprising baseline, mental stress, recovery from the stress, relaxation, and recovery from the relaxation, respectively. For each phase, the mean amplitude of the skin conductance level (MSCL), standard deviations of the SCL (SDSCL), slope of the SCL (SSCL), mean amplitude of the non-specific skin conductance responses (MSCR), number of non-specific skin conductance responses (NSCR), and power spectral density (PSD) were evaluated from the SC signals, producing 30 parameters overall (six features for each phase). These features were used as input data for a support vector machine (SVM) algorithm designed to distinguish MDD patients from healthy controls based on their physiological responses. Statistical tests showed that the main effect of task was significant in all SC features, and the main effect of group was significant in MSCL, SDSCL, SSCL, and PSD. In addition, the proposed algorithm achieved 70% accuracy, 70% sensitivity, 71% specificity, 70% positive predictive value, 71% negative predictive value in classifying MDD patients and healthy controls. These results demonstrated that it is possible to extract meaningful features that reflect changes in ANS responses to various stimuli. Using these features, detection of MDD was feasible, suggesting that SC analysis has great potential for future diagnostics and prediction of depression based on objective interpretation of depressive states. [ABSTRACT FROM AUTHOR]

Although successful detection of malignant tumors from histopathological images largely depends on the long-term experience of radiologists, experts sometimes disagree with their decisions. Computer-aided diagnosis provides a second option for image diagnosis, which can improve the reliability of experts’ decision-making. Automatic and precision classification for breast cancer histopathological image is of great importance in clinical application for identifying malignant tumors from histopathological images. Advanced convolution neural network technology has achieved great success in natural image classification, and it has been used widely in biomedical image processing. In this paper, we design a novel convolutional neural network, which includes a convolutional layer, small SE-ResNet module, and fully connected layer. We propose a small SE-ResNet module which is an improvement on the combination of residual module and Squeeze-and-Excitation block, and achieves the similar performance with fewer parameters. In addition, we propose a new learning rate scheduler which can get excellent performance without complicatedly fine-tuning the learning rate. We use our model for the automatic classification of breast cancer histology images (BreakHis dataset) into benign and malignant and eight subtypes. The results show that our model achieves the accuracy between 98.87% and 99.34% for the binary classification and achieve the accuracy between 90.66% and 93.81% for the multi-class classification. [ABSTRACT FROM AUTHOR]

Chronic Respiratory Diseases (CRDs), such as Asthma and Chronic Obstructive Pulmonary Disease (COPD), are leading causes of deaths worldwide. Although both Asthma and COPD are not curable, they can be managed by close monitoring of symptoms to prevent worsening of the condition. One key symptom that needs to be monitored is the occurrence of wheezing sounds during breathing since its early identification could prevent serious exacerbations. Since wheezing can happen randomly without warning, a long-term monitoring system with automatic wheeze detection could be extremely helpful to manage these respiratory diseases. This study evaluates the discriminatory ability of different types of feature used in previous related studies, with a total size of 105 individual features, for automatic identification of wheezing sound during breathing. A linear classifier is used to determine the best features for classification by evaluating several performance metrics, including ranksum statistical test, area under the sensitivity-–specificity curve (AUC), F1 score, Matthews Correlation Coefficient (MCC), and relative computation time. Tonality index attained the highest effect size, at 87.95%, and was found to be the feature with the lowest p-value when ranksum significance test was performed. Third MFCC coefficient achieved the highest AUC and average optimum F1 score at 0.8919 and 82.67% respectively, while the highest average optimum MCC was obtained by the first coefficient of a 6th order LPC. The best possible combination of two and three features for wheeze detection is also studied. The study concludes with an analysis of the different trade-offs between accuracy, reliability, and computation requirements of the different features since these will be highly useful for researchers when designing algorithms for automatic wheeze identification. [ABSTRACT FROM AUTHOR]