Search

Diabetes intensive care unit (ICU) patients are at increased risk of complications leading to in-hospital mortality. Assessing the likelihood of death is a challenging and time-consuming task due to a large number of influencing factors. Healthcare providers are interested in the detection of ICU patients at higher risk, such that risk factors can possibly be mitigated. While such severity scoring methods exist, they are commonly based on a snapshot of the health conditions of a patient during the ICU stay and do not specifically consider a patient's prior medical history. In this paper, a process mining/deep learning architecture is proposed to improve established severity scoring methods by incorporating the medical history of diabetes patients. First, health records of past hospital encounters are converted to event logs suitable for process mining. The event logs are then used to discover a process model that describes the past hospital encounters of patients. An adaptation of Decay Replay Mining is proposed to combine medical and demographic information with established severity scores to predict the in-hospital mortality of diabetes ICU patients. Significant performance improvements are demonstrated compared to established risk severity scoring methods and machine learning approaches using the Medical Information Mart for Intensive Care III dataset.

Classification models for multivariate time series have drawn the interest of many researchers to the field with the objective of developing accurate and efficient models. However, limited research has been conducted on generating adversarial samples for multivariate time series classification models. Adversarial samples could become a security concern in systems with complex sets of sensors. This study proposes extending the existing gradient adversarial transformation network (GATN) in combination with adversarial autoencoders to attack multivariate time series classification models. The proposed model attacks classification models by utilizing a distilled model to imitate the output of the multivariate time series classification model. In addition, the adversarial generator function is replaced with a variational autoencoder to enhance the adversarial samples. The developed methodology is tested on two multivariate time series classification models: 1-nearest neighbor dynamic time warping (1-NN DTW) and a fully convolutional network (FCN). This study utilizes 30 multivariate time series benchmarks provided by the University of East Anglia (UEA) and University of California Riverside (UCR). The use of adversarial autoencoders shows an increase in the fraction of successful adversaries generated on multivariate time series. To the best of our knowledge, this is the first study to explore adversarial attacks on multivariate time series. Additionally, we recommend future research utilizing the generated latent space from the variational autoencoders.

BackgroundParalytic Ileus (PI) patients in the Intensive Care Unit (ICU) are at a significant risk of death. Prediction of at-risk patients for mortality after 24 hours of admission of ICU PI patients is important to increase the life expectancy of PI patients.Methods and ResultsThe proposed framework,DLMP(Deep Learning Model for Mortality Prediction of ICU Patients with PI) is a powerful deep learning model consisting of six total unique clinical lab items and two demographics as inputs to a Neural Network(NN) of only two neuron layers. Using the Medical Information Mart for Intensive Care III (MIMIC-III) dataset of 1,017 ICU PI patients, theDLMPresulted in the best prediction performance with an AUC score of 0.866.ConclusionThe proposed approach is capable of modeling the mortality of ICU patients after 24 hours admission using only six unique total clinical data and two demographics with a simple NN architecture.DLMPframework significantly improves the outcome prediction compared to the process mining and machine learning models. The proposedDLMPhas the potential of allowing clinicians to create targeted interventions that reduce mortality for PI patients in an ICU setting.

In process mining, process models are extracted from event logs using process discovery algorithms and are commonly assessed using multiple quality dimensions. While the metrics that measure the relationship of an extracted process model to its event log are well-studied, quantifying the level by which a process model can describe the unobserved behavior of its underlying system falls short in the literature. In this paper, a novel deep learning-based methodology called Adversarial System Variant Approximation (AVATAR) is proposed to overcome this issue. Sequence Generative Adversarial Networks are trained on the variants contained in an event log with the intention to approximate the underlying variant distribution of the system behavior. Unobserved realistic variants are sampled either directly from the Sequence Generative Adversarial Network or by leveraging the Metropolis-Hastings algorithm. The degree by which a process model relates to its underlying unknown system behavior is then quantified based on the realistic observed and estimated unobserved variants using established process model quality metrics. Significant performance improvements in revealing realistic unobserved variants are demonstrated in a controlled experiment on 15 ground truth systems. Additionally, the proposed methodology is experimentally tested and evaluated to quantify the generalization of 60 discovered process models with respect to their systems.

Background Out-of-hospital cardiac arrest (OHCA) affects nearly 400,000 people each year in the United States of which only 10% survive. Using data from the Cardiac Arrest Registry to Enhance Survival (CARES), and machine learning (ML) techniques, we developed a model of neurological outcome prediction for OHCA in Chicago, Illinois. Methods Rescue workflow data of 2639 patients with witnessed OHCA were retrieved from Chicago’s CARES. An Embedded Fully Convolutional Network (EFCN) classification model was selected to predict the patient outcome (survival with good neurological outcomes or not) based on 27 input features with the objective of maximizing the average class sensitivity. Using this model, sensitivity analysis of intervention variables such as bystander cardiopulmonary resuscitation (CPR), targeted temperature management, and coronary angiography was conducted. Results The EFCN classification model has an average class sensitivity of 0.825. Sensitivity analysis of patient outcome shows that an additional 33 patients would have survived with good neurological outcome if they had received lay person CPR in addition to CPR by emergency medical services and 88 additional patients would have survived if they had received the coronary angiography intervention. Conclusions ML modeling of the complex Chicago OHCA rescue system can predict neurologic outcomes with a reasonable level of accuracy and can be used to support intervention decisions such as CPR or coronary angiography. The discriminative ability of this ML model requires validation in external cohorts to establish generalizability.

Introduction: The field of artificial intelligence (AI) is rapidly expanding, with many applications seen routinely in health care, industry, and education, and increasingly in workplaces. Although there is growing evidence of applications of AI in workplaces across all industries to simplify and/or automate tasks there is a limited understanding of the role that AI contributes in addressing occupational safety and health (OSH) concerns. Methods: This paper introduces a new framework called Risk Evolution, Detection, Evaluation, and Control of Accidents (REDECA) that highlights the role that AI plays in the anticipation and control of exposure risks in a worker’s immediate environment. Two hundred and sixty AI papers across five sectors (oil and gas, mining, transportation, construction, and agriculture) were reviewed using the REDECA framework to highlight current applications and gaps in OSH and AI fields. Results: The REDECA framework highlighted the unique attributes and research focus of each of the five industrial sectors. The majority of evidence of AI in OSH research within the oil/gas and transportation sectors focused on the development of sensors to detect hazardous situations. In construction the focus was on the use of sensors to detect incidents. The research in the agriculture sector focused on sensors and actuators that removed workers from hazardous conditions. Application of the REDECA framework highlighted AI/OSH strengths and opportunities in various industries and potential areas for collaboration. Conclusions: As AI applications across industries continue to increase, further exploration of the benefits and challenges of AI applications in OSH is needed to optimally protect worker health, safety and well-being.

Time series classification models have been garnering significant importance in the research community. However, not much research has been done on generating adversarial samples for these models. These adversarial samples can become a security concern. In this paper, we propose utilizing an adversarial transformation network (ATN) on a distilled model to attack various time series classification models. The proposed attack on the classification model utilizes a distilled model as a surrogate that mimics the behavior of the attacked classical time series classification models. Our proposed methodology is applied onto 1-Nearest Neighbor Dynamic Time Warping (1-NN ) DTW, a Fully Connected Network and a Fully Convolutional Network (FCN), all of which are trained on 42 University of California Riverside (UCR) datasets. In this paper, we show both models were susceptible to attacks on all 42 datasets. To the best of our knowledge, such an attack on time series classification models has never been done before. Finally, we recommend future researchers that develop time series classification models to incorporating adversarial data samples into their training data sets to improve resilience on adversarial samples and to consider model robustness as an evaluative metric., 13 pages, 7 figures, 6 tables

Long Short Term Memory Fully Convolutional Neural Networks (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN) have shown to achieve state-of-the-art performance on the task of classifying time series signals on the old University of California-Riverside (UCR) time series repository. However, there has been no study on why LSTM-FCN and ALSTM-FCN perform well. In this paper, we perform a series of ablation tests (3627 experiments) on LSTM-FCN and ALSTM-FCN to provide a better understanding of the model and each of its sub-module. Results from the ablation tests on ALSTM-FCN and LSTM-FCN show that the LSTM and the FCN blocks perform better when applied in a conjoined manner. Two z-normalizing techniques, z-normalizing each sample independently and z-normalizing the whole dataset, are compared using a Wilcoxson signed-rank test to show a statistical difference in performance. In addition, we provide an understanding of the impact dimension shuffle has on LSTM-FCN by comparing its performance with LSTM-FCN when no dimension shuffle is applied. Finally, we demonstrate the performance of the LSTM-FCN when the LSTM block is replaced by a GRU, basic RNN, and Dense Block., 8 pages, 4 tables, 1 figure

Dynamic Time Warping (DTW) is a popular method for measuring the similarity of time series. It is widely used in various domains. A major drawback of DTW is that it has a high computational complexity. To address this problem, pruning techniques to calculate the exact DTW distance, as well as DTW approximation methods, have become important approaches. In this paper, we introduce Blocked Dynamic Time Warping (BDTW), a new similarity measure which works on run-length encoded time series representation. BDTW utilizes any repetitive values (zero and nonzero) in time series to reduce DTW computation time. BDTW closely approximates DTW distance, and it is significantly faster than traditional DTW for time series with high levels of value repetition. Moreover, BDTW can be combined with time series representation methods which provide constant segments, to serve as a close approximation method even for the time series without value repetition. Constrained BDTW, BDTW upper bound and BDTW lower bound are discussed as variations of BDTW. BDTW upper bound and BDTW lower bound are presented as a new DTW upper bound and lower bound which can be efficiently applied on time series with high levels of value repetition for pruning unhopeful alignments and matches in the exact DTW calculation. We show the effectiveness of BDTW and its variations on different applications using the following datasets: Almanac of Minutely Power, Refit Smart Homes, as well as the 85 datasets from the University of California, Riverside time series classification archive (UCR archive).

Fully convolutional neural networks (FCN) have been shown to achieve state-of-the-art performance on the task of classifying time series sequences. We propose the augmentation of fully convolutional networks with long short term memory recurrent neural network (LSTM RNN) sub-modules for time series classification. Our proposed models significantly enhance the performance of fully convolutional networks with a nominal increase in model size and require minimal preprocessing of the dataset. The proposed Long Short Term Memory Fully Convolutional Network (LSTM-FCN) achieves state-of-the-art performance compared to others. We also explore the usage of attention mechanism to improve time series classification with the Attention Long Short Term Memory Fully Convolutional Network (ALSTM-FCN). Utilization of the attention mechanism allows one to visualize the decision process of the LSTM cell. Furthermore, we propose fine-tuning as a method to enhance the performance of trained models. An overall analysis of the performance of our model is provided and compared to other techniques., 7 pages, 3 figures and 2 tables

Throughout recent years, dynamic time warping (DTW) has remained as a robust similarity measure in time series classification (TSC). 1-nearest neighbor (1-NN) algorithm with DTW is the most widely used classification method on time series serving as a benchmark. With the increasing demand for TSC on low-resource devices and the widespread of wearable devices, the need for a efficient and accurate time series classifier has never been higher. Although 1-NN DTW attains accurate results, it highly falls back on efficiency due to its quadratic complexity in the length of time series. In this paper, we propose a new approximation method for reducing the length of the time series as the input of DTW. We call it control chart approximation (CCA), after a similar concept used in statistical quality control processing. CCA representation approximates raw time series by transforming them into a set of segments with aggregated values and durations forming a reduced 3-D vector. We also propose an adaptation of DTW in 3-D space as a distance measure for 1-NN classifier, and denote the method as 1-NN 3-D DTW. Our experiments on 85 datasets from UCR archive—including 28 long-length (>500 points) time series datasets—show up to two orders of magnitude performance gain in running time compared to the state-of-the-art 1-NN DTW implementation. Moreover, it shows similar or better accuracy on the long time series in the experiment.

Determining the optimal initial dose for warfarin is a critically important task. Several factors have an impact on the therapeutic dose for individual patients, such as patients' physical attributes (Age, Height, etc.), medication profile, co-morbidities, and metabolic genotypes (CYP2C9 and VKORC1). These wide range factors influencing therapeutic dose, create a complex environment for clinicians to determine the optimal initial dose. Using a sample of 4,237 patients, we have proposed a companion classification model to one of the most popular dosing algorithms (International Warfarin Pharmacogenetics Consortium (IWPC) clinical model), which identifies the appropriate cohort of patients for applying this model. The proposed model functions as a clinical decision support system which assists clinicians in dosing. We have developed a classification model using Support Vector Machines, with a polynomial kernel function to determine if applying the dose prediction model is appropriate for a given patient. The IWPC clinical model will only be used if the patient is classified as "Safe for model". By using the proposed methodology, the dosing mode's prediction accuracy increases by 15 percent in terms of Root Mean Squared Error and 17 percent in terms of Mean Absolute Error in dose estimates of patients classified as "Safe for model"., Comment: 6 pages, 8 tables, 1 figure

Over the past decade, multivariate time series classification has received great attention. We propose transforming the existing univariate time series classification models, the Long Short Term Memory Fully Convolutional Network (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN), into a multivariate time series classification model by augmenting the fully convolutional block with a squeeze-and-excitation block to further improve accuracy. Our proposed models outperform most state-of-the-art models while requiring minimum preprocessing. The proposed models work efficiently on various complex multivariate time series classification tasks such as activity recognition or action recognition. Furthermore, the proposed models are highly efficient at test time and small enough to deploy on memory constrained systems., 18 pages, 2 figures, 5 tables

Time series data are generated in abundance of information technology applications. Classification of time series is one of the significant areas of interest in Time Series Data Mining. Over the last decade, different approaches have been developed to achieve a superior effectiveness and efficiency in classification of time series. We propose a new model for classification of time series which first uses SAX (symbolic aggregate approximation) method to transform the time series to symbols and then applies a probabilistic methodology in dealing with classification of symbolic sequences. This model is developed based on Bayesian rule and Probability Chain rule on symbolic representation of time series along with considering a penalty for mismatching points. We call this model Bayesian and Chain rule Model (BCM). The performance of BCM has been evaluated using 43 time series data sets. We performed an extensive experiment to compare the performance of BCM to Euclidean Distance, Dynamic Time Warping (DTW) with best warping window, and DTW with no warping window. This comparison was done through three stages. Stage 1 was predicting the domains and problems, in which BCM works relatively well. Stage 2 included measuring the prediction accuracy of BCM to those of the other methods. Stage 3 involved constructing the confusion matrixes for predictions in Stage 1 as well as measuring and comparing the predictability of the methods. The results show that, compared to the other popular time series classification methods, BCM is more accurate where it is predicted to perform better.

This paper describes optimization of a sensor network by a Genetic Algorithm (GA). The system developed automatically generates the optimization problems depending on the events happening in the environment and constructs a GA with the appropriate internal structure for the problem at hand. GA finds the quasioptimal combination of sensors that can detect and/or locate the targets. An optimal combination is the one that minimizes the power consumption of the entire sensor network and gives the best accuracy of location of desired targets. The paper attempts to determine the percentage of the total search space (PTSS) that should be covered by GA in order to obtain consistent quasioptimal solutions in different runs. The second goal is to determine the relationship between the population size and the GA stopping criteria that for a given PTSS ensures the best performance of GA. The study is performed for the sensor network optimization problem with three targets.

This paper proposes a new model for predicting the optimal warfarin dosing for African American patients. The prediction model is created using the multivariable regression method. The accuracy of dosing prediction is directly related to patient's safety. We show that the proposed model has better accuracy compare to all other available prediction methods for optimal dosing of warfarin.

In this paper, we develop a framework for reconfiguration of a discrete event system (DES) controller, which has a dynamic event observation set. We will show the designed reconfiguration yields a more tolerable controller than the one designed in [5]. Starting with a maximally permissive controller that has a full observation of its DES, we design a mega-controller that monitors the observation set of the DES controller and its state continuously. Upon a change in the observation set, the mega-controller reconfigures the controller by a aggregation or disaggregation of the controller states. The mega-controller is also responsible for feedback function adjustments if the available observation set causes a conflict in control. We illustrate the reconfiguration procedures by an example.