Prediction Of Critical Events Research Articles

Development of a deep learning model that predicts critical events of pediatric patients admitted to general wards

Early detection of deteriorating patients is important to prevent life-threatening events and improve clinical outcomes. Efforts have been made to detect or prevent major events such as cardiopulmonary resuscitation, but previously developed tools are often complicated and time-consuming, rendering them impractical. To overcome this problem, we designed this study to create a deep learning prediction model that predicts critical events with simplified variables. This retrospective observational study included patients under the age of 18 who were admitted to the general ward of a tertiary children’s hospital between 2020 and 2022. A critical event was defined as cardiopulmonary resuscitation, unplanned transfer to the intensive care unit, or mortality. The vital signs measured during hospitalization, their measurement intervals, sex, and age were used to train a critical event prediction model. Age-specific z-scores were used to normalize the variability of the normal range by age. The entire dataset was classified into a training dataset and a test dataset at an 8:2 ratio, and model learning and testing were performed on each dataset. The predictive performance of the developed model showed excellent results, with an area under the receiver operating characteristics curve of 0.986 and an area under the precision-recall curve of 0.896. We developed a deep learning model with outstanding predictive power using simplified variables to effectively predict critical events while reducing the workload of medical staff. Nevertheless, because this was a single-center trial, no external validation was carried out, prompting further investigation.

Machine learning driven smart fire safety design of false ceiling and emergency response

In modern buildings, false ceilings are widely used for building services systems and aesthetic purposes, but they also pose challenges in terms of fire safety. A fire accident typically results from the failure of multiple safety measures or components. In many fire accidents, fire and smoke reached the false ceiling and kept spreading in the interstitial space without any detection. This study first generates a numerical database of false ceiling fire scenarios by varying the room dimensions, false-ceiling leakage area, fire size and locations. Then, a smart model based on machine learning to predict the fire smoke motion below and above the false ceiling is developed. The trained model is capable to predict the activation time of fire detectors and sprinklers for any given false ceiling design and fire scenario. This methodology enables a designer to generate multiple fire scenarios and determine the available safe egress time (ASET) for performance-based fire engineering designs especially in terms of fire detection time. In case of a real fire, with the data feed from the fire sensor network, the trained machine learning model can further predict the critical building fire events with the false ceiling, such as multi-compartment fires, smoke in the evacuation path, and structural failures. This work proposes a smart framework for improving the building fire safety design of false ceilings and the sensor-driven fire forecast to support firefighting.It enhances the emergency response processes by enabling dynamic risk assessment through prediction of critical events.

Prediction of sports injuries by psychological process monitoring

Abstract: Objectives: Sports injuries usually have severe consequences for the concerned athletes as well as for trainers and teams. The question is if accidents can be predicted in specific cases. Can early-warning signals be detected in psychological time series? Methods: An App-based method of process-monitoring was applied for data collection of psychological parameters. Daily self-assessments using a Sports Process Questionnaire were realized by a professional soccer player during the after-care period of a psychiatric treatment. Methods for the prediction of critical events were applied (Dynamic Complexity, Recurrence Plots, dynamic inter-item correlations). Injuries may demarcate pattern transitions in the mental functioning of athletes, which could be identified by the Pattern Transition Detection Algorithm (PTDA). Results: Early-warning signals of the accident could be identified in the time series. Dynamic Complexity revealed a critical instability, Recurrence Plots a transient period, and the dynamic inter-item correlations a period of increased system coherence just before the accident. The PTDA revealed a phase transition at the occurring injury. Conclusions: Even if the analysis is based on a single case, the results are promising. Psychological self-reports allow a short-term prediction of bio-mechanical injuries and by this, can help to prevent them. Nonlinear measures can be applied to time series data collected by digital process monitoring.

Contrastive learning improves critical event prediction in COVID-19 patients

SummaryDeep learning (DL) models typically require large-scale, balanced training data to be robust, generalizable, and effective in the context of healthcare. This has been a major issue for developing DL models for the coronavirus disease 2019 (COVID-19) pandemic, where data are highly class imbalanced. Conventional approaches in DL use cross-entropy loss (CEL), which often suffers from poor margin classification. We show that contrastive loss (CL) improves the performance of CEL, especially in imbalanced electronic health records (EHR) data for COVID-19 analyses. We use a diverse EHR dataset to predict three outcomes: mortality, intubation, and intensive care unit (ICU) transfer in hospitalized COVID-19 patients over multiple time windows. To compare the performance of CEL and CL, models are tested on the full dataset and a restricted dataset. CL models consistently outperform CEL models, with differences ranging from 0.04 to 0.15 for area under the precision and recall curve (AUPRC) and 0.05 to 0.1 for area under the receiver-operating characteristic curve (AUROC).

Acquisition of Semantics for AI-based Applications in Medical Technologies

Abstract For the development, training, and validation of AIbased procedures, such as the analysis of clinical data, prediction of critical events, or planning of healthcare procedures, a lot of data is needed. In addition to this data of any origin (image data, bio-signals, health records, machine states, …) adequate supplementary information about the meaning encoded in the data is required. With this additional information - the semantic or knowledge - a tight relation between the raw data and the human-understandable concepts from the real world can be established. Nevertheless, as the amount of data needed to develop robust AI-based methods is strongly increasing, the assessment and acquisition of the related knowledge becomes more and more challenging. Within this work, an overview of currently available concepts of knowledge acquisition are described and evaluated. Four main groups of knowledge acquisition related to AI-based technologies have been identified. For image data mainly iconic annotation methods are used, where experienced users mark or draw depicted entities in the images and label them using predefined sets of classifications. Similarly, bio-signals are manually labelled, whereby important events along the timeline are marked. If no sufficient data is available, augmentation and simulation techniques are applied yielding data and semantics at the same time. In applications, where expensive sensors are replaced by low-cost devices, the high-grade data can be used as semantics. Finally, classic rule-based approaches are used, where human factual and procedural knowledge about the data and its context is translated into machine-understandable procedures. All these methods are depending on the involvement of human experts. To reduce this, more intelligent and hybrid approaches are needed, shifting the focus from the-human-in-the-loop to the-machine- in-the-loop.

SARS-CoV-2 Epidemics in Retirement and Nursing Homes in Italy: A New Preparedness Assessment Model after the First Epidemic Wave.

The aim of the study is to evaluate the preparedness of retirement and nursing homes in the city of Sassari at the end of the first wave of the severe acute respiratory syndrome coronavirus 2 epidemic, first by investigating the risk perception of epidemic outbreaks by the facility managers and subsequently by carrying out a field assessment of these facilities. To perform the field assessment, a checklist developed by the CDC (Infection Prevention and Control Assessment Tool for Nursing Homes Preparing for COVID-19) and adapted to the Italian context was used. Fourteen facilities took part in the survey (87.5%). The application of good practices for each survey area was expressed as a percentage with the following median values: restriction policies (87.5%), staff training (53.8%), resident training (67.6%), availability of personal protective equipment (41.7%), infection control practices (73.5%) and communication (80%). Among the facilities, considerable variability was observed in these evaluation fields: only the restriction policies and communication activities were applied uniformly. A discrepancy was found between perceived risk and real danger in the facilities, requiring targeted communication actions. At present, it is necessary to promote a new approach based on the prediction of critical events, thereby providing the means to effectively address them.

Safety critical event prediction through unified analysis of driver and vehicle volatilities: Application of deep learning methods

Transportation safety is highly correlated with driving behavior, especially human error playing a key role in a large portion of crashes. Modern instrumentation and computational resources allow for the monitorization of driver, vehicle, and roadway/environment to extract leading indicators of crashes from multi-dimensional data streams. To quantify variations that are beyond normal in driver behavior and vehicle kinematics, the concept of volatility is applied. The study measures driver-vehicle volatilities using the naturalistic driving data. By integrating and fusing multiple real-time streams of data, i.e., driver distraction, vehicular movements and kinematics, and instability in driving, this study aims to predict occurrence of safety critical events and generate appropriate feedback to drivers and surrounding vehicles. The naturalistic driving data is used which contains 7566 normal driving events, and 1315 severe events (i.e., crash and near-crash), vehicle kinematics, and driver behavior collected from more than 3500 drivers. In order to capture the local dependency and volatility in time-series data 1D-Convolutional Neural Network (1D-CNN), Long Short-Term Memory (LSTM), and 1DCNN-LSTM are applied. Vehicle kinematics, driving volatility, and impaired driving (in terms of distraction) are used as the input parameters. The results reveal that the 1DCNN-LSTM model provides the best performance, with 95.45% accuracy and prediction of 73.4% of crashes with a precision of 95.67%. Additional features are extracted with the CNN layers and temporal dependency between observations is addressed, which helps the network learn driving patterns and volatile behavior. The model can be used to monitor driving behavior in real-time and provide warnings and alerts to drivers in low-level automated vehicles, reducing their crash risk.

Machine Learning to Predict Mortality and Critical Events in a Cohort of Patients With COVID-19 in New York City: Model Development and Validation.

BackgroundCOVID-19 has infected millions of people worldwide and is responsible for several hundred thousand fatalities. The COVID-19 pandemic has necessitated thoughtful resource allocation and early identification of high-risk patients. However, effective methods to meet these needs are lacking.ObjectiveThe aims of this study were to analyze the electronic health records (EHRs) of patients who tested positive for COVID-19 and were admitted to hospitals in the Mount Sinai Health System in New York City; to develop machine learning models for making predictions about the hospital course of the patients over clinically meaningful time horizons based on patient characteristics at admission; and to assess the performance of these models at multiple hospitals and time points.MethodsWe used Extreme Gradient Boosting (XGBoost) and baseline comparator models to predict in-hospital mortality and critical events at time windows of 3, 5, 7, and 10 days from admission. Our study population included harmonized EHR data from five hospitals in New York City for 4098 COVID-19–positive patients admitted from March 15 to May 22, 2020. The models were first trained on patients from a single hospital (n=1514) before or on May 1, externally validated on patients from four other hospitals (n=2201) before or on May 1, and prospectively validated on all patients after May 1 (n=383). Finally, we established model interpretability to identify and rank variables that drive model predictions.ResultsUpon cross-validation, the XGBoost classifier outperformed baseline models, with an area under the receiver operating characteristic curve (AUC-ROC) for mortality of 0.89 at 3 days, 0.85 at 5 and 7 days, and 0.84 at 10 days. XGBoost also performed well for critical event prediction, with an AUC-ROC of 0.80 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. In external validation, XGBoost achieved an AUC-ROC of 0.88 at 3 days, 0.86 at 5 days, 0.86 at 7 days, and 0.84 at 10 days for mortality prediction. Similarly, the unimputed XGBoost model achieved an AUC-ROC of 0.78 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. Trends in performance on prospective validation sets were similar. At 7 days, acute kidney injury on admission, elevated LDH, tachypnea, and hyperglycemia were the strongest drivers of critical event prediction, while higher age, anion gap, and C-reactive protein were the strongest drivers of mortality prediction.ConclusionsWe externally and prospectively trained and validated machine learning models for mortality and critical events for patients with COVID-19 at different time horizons. These models identified at-risk patients and uncovered underlying relationships that predicted outcomes.

Risk Assessment and Prediction of Severe or Critical COVID-19 Illness in Older Adults.

PurposeThis study is to investigate the risk prediction of severe or critical events of COVID-19 in older adults in China and provide the evidence to support the management of older adults with COVID-19.Materials and MethodsThe clinical data of older adults with COVID-19 admitted to the Shanghai Public Health Clinical Center during January 20, 2020 to March 16, 2020 were collected. The possible risk factors of severe or critical illness were investigated with Cox proportional hazards (PH) regression models for univariate and multivariate analyses to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). For the prediction indicators, optimum cut-off points were determined by calculating Youden’s index. The efficacy of risk prediction of severe or critical illness was examined through the receiver operating characteristic (ROC) curve.ResultsA total of 110 older adults with COVID-19 were included, in which 21 (19.1%) patients had severe or critical illness of COVID-19. Multivariable regression analysis showed that CD4 cells and D-dimer were independent risk factors. D-dimer, CD4 cells, and CD cells/D-dimer ratio with cut-off values of 0.65 (mg/L), 268 (cell/µL) and 431 were in the prediction of severe or critical illness of older adults with COVID-19. The AUC value of D-dimer, CD4 cells, CD4 cells/D-dimer ratio, the tandem combination and the parallel combination to predict severe or critical illness of the older adults with COVID-19 were 0.703, 0.804, 0.794, 0.812 and 0.694, respectively.ConclusionD-dimer and CD4 cells either by themselves or in combination have demonstrated predictive value in risk stratification as well as established the prognosis of severe or critical illness in older adults with COVID-19.

Developing a novel force forecasting technique for early prediction of critical events in robotics.

Safety critical events in robotic applications can often be characterized by forces between the robot end-effector and the environment. One application in which safe interaction between the robot and environment is critical is in the area of medical robots. In this paper, we propose a novel Compact Form Dynamic Linearization Model-Free Prediction (CFDL-MFP) technique to predict future values of any time-series sensor data, such as interaction forces. Existing time series forecasting methods have high computational times which motivates the development of a novel technique. Using Autoregressive Integrated Moving Average (ARIMA) forecasting as benchmark, the performance of the proposed model was evaluated in terms of accuracy, computation efficiency, and stability on various force profiles. The proposed algorithm was 11% more accurate than ARIMA and maximum computation time of CFDL-MFP was 4ms, compared to ARIMA (7390ms). Furthermore, we evaluate the model in the special case of predicting needle buckling events, before they occur, by using only axial force and needle-tip position data. The model was evaluated experimentally for robustness with steerable needle insertions into different tissues including gelatin and biological tissue. For a needle insertion velocity of 2.5mm/s, the proposed algorithm was able to predict needle buckling 2.03s sooner than human detections. In biological tissue, no false positive or false negative buckling detections occurred and the rates were low in artificial tissue. The proposed forecasting model can be used to ensure safe robot interactions with delicate environments by predicting adverse force-based events before they occur.

Modeling and Predicting Community Structure Changes in Time-Evolving Social Networks

As time evolves, communities in a social network may undergo various changes known as critical events. For instance, a community can either split into several other communities, expand into a larger community, shrink to a smaller community, remain stable or merge into another community. Prediction of critical events has attracted increasing attention in the recent literature. Learning the evolution of communities over time is a key step towards predicting the critical events the communities may undergo. This is an important and difficult issue in the study of social networks. In the work to date, there is a lack of formal approaches for modeling and predicting critical events over time. This motivates our effort to design a new statistical method for event prediction in order to make better use of histories of past changes. To this end, this paper proposes a sliding window analysis from which we develop a model that simultaneously exploits an autoregressive model and survival analysis techniques. The autoregressive model is employed here to simulate the evolution of the community structure, whereas the survival analysis techniques allow the prediction of future changes the community may undergo.

Early prediction of critical events for infants with single-ventricle physiology in critical care using routinely collected data

ObjectiveCritical events are common and difficult to predict among infants with congenital heart disease and are associated with mortality and long-term sequelae. We aimed to achieve early prediction of critical events, that is, cardiopulmonary resuscitation, emergency endotracheal intubation, and extracorporeal membrane oxygenation in infants with single-ventricle physiology before second-stage surgery. We hypothesized that naïve Bayesian models learned from expert knowledge and clinical data can predict critical events early and accurately. MethodsWe collected 93 patients with single-ventricle physiology admitted to intensive care units in a single tertiary pediatric hospital between 2014 and 2017. Using knowledge elicited from experienced cardiac-intensive-care-unit providers and machine-learning techniques, we developed and evaluated the Cardiac-intensive-care Warning INdex (C-WIN) system, consisting of a set of naïve Bayesian models that leverage routinely collected data. We evaluated predictive performance using the area under the receiver operating characteristic curve, sensitivity, and specificity. We performed the evaluation at 5 different prediction horizons: 1, 2, 4, 6, and 8 hours before the onset of critical events. ResultsThe area under the receiver operating characteristic curves of the C-WIN models ranged between 0.73 and 0.88 at different prediction horizons. At 1 hour before critical events, C-WIN was able to detect events with an area under the receiver operating characteristic curve of 0.88 (95% confidence interval, 0.84-0.92) and a sensitivity of 84% at the 81% specificity level. ConclusionsPredictive models may enhance clinicians' ability to identify infants with single-ventricle physiology at high risk of critical events. Early prediction of critical events may indicate the need to perform timely interventions, potentially reducing morbidity, mortality, and health care costs.

Innovative Monitoring Tools and Early Warning Systems for Risk Management: A Case Study

During recent years, the availability of innovative monitoring instrumentation has been a fundamental component in the development of efficient and reliable early warning systems (EWS). In fact, the potential to achieve high sampling frequencies, together with automatic data transmission and elaboration are key features for a near-real time approach. This paper presents a case study located in Central Italy, where the realization of an important state route required a series of preliminary surveys. The monitoring system installed on site included manual inclinometers, automatic modular underground monitoring system (MUMS) inclinometers, piezometers, and geognostic surveys. In particular, data recorded by innovative instrumentation allowed for the detection of major slope displacements that ultimately led to the landslide collapse. The implementation of advanced tools, featuring remote and automatic procedures for data sampling and elaboration, played a key role in the critical event identification and prediction. In fact, thanks to displacement data recorded by the MUMS inclinometer, it was possible to forecast the slope failure that was later confirmed during the following site inspection. Additionally, a numerical analysis was performed to better understand the mechanical behavior of the slope, back-analyze the monitored event, and to assess the stability conditions of the area of interest.

Multivariate Time-Series Similarity Assessment via Unsupervised Representation Learning and Stratified Locality Sensitive Hashing: Application to Early Acute Hypotensive Episode Detection

An accurate and timely prediction of clinically critical events in intensive care unit (ICU) is important for improving care and survival rate. Most of the existing approaches are based on the application of various classification methods on explicitly extracted statistical features from the vital signals of ICU patients, such as the mean, the standard deviation, and the skew. In this article, we propose to eliminate the high cost of engineering hand-crafted features from multivariate time-series of physiologic signals by learning its representation with a sequence-to-sequence autoencoder. We then propose to hash the learned multivariate time-series representations of labeled dataset to enable signal similarity assessment for the prediction of critical events. We evaluate this methodological framework to predict acute hypotensive episodes (AHE) on a large and diverse dataset of vital signal recordings extracted from eICU collaborative research database. Experiments demonstrate the ability of the presented framework in accurately predicting an upcoming AHE.

Adverse condition and critical event prediction in commercial buildings: Danish case study

Over the last two decades, there has been a growing realization that the actual energy performances of many buildings fail to meet the original intent of building design. Faults in systems and equipment, incorrectly configured control systems and inappropriate operating procedures increase the energy consumption about 20% and therefore compromise the building energy performance. To improve the energy performance of buildings and to prevent occupant discomfort, adverse condition and critical event prediction plays an important role. The Adverse Condition and Critical Event Prediction Toolbox (ACCEPT) is a generic framework to compare and contrast methods that enable prediction of an adverse event, with low false alarm and missed detection rates. In this paper, ACCEPT is used for fault detection and prediction in a real building at the University of Southern Denmark. To make fault detection and prediction possible, machine learning methods such as Kernel Density Estimation (KDE), and Principal Component Analysis (PCA) are used. A new PCA–based method is developed for artificial fault generation. While the proposed method finds applications in different areas, it has been used primarily for analysis purposes in this work. The results are evaluated, discussed and compared with results from Canonical Variate Analysis (CVA) with KDE. The results show that ACCEPT is more powerful than CVA with KDE which is known to be one of the best multivariate data-driven techniques in particular, under dynamically changing operational conditions.

Collision frequency locality-sensitive hashing for prediction of critical events.

We present a fast, efficient method to predict future critical events for a patient. The prediction method is based on retrieving and leveraging similar waveform trajectories from a large medical database. Locality-sensitive hashing (LSH), our theoretical foundation, is a model-free, sub-linear time, approximate search method enabling a fast retrieval of a nearest neighbor set for a given query. We propose a new variant of LSH, namely Collision Frequency LSH (CFLSH), to further improve the prediction accuracy without sacrificing any speed. The key idea is that the more frequently an element and a query collide across multiple LSH hash tables, the more similar they are. Unlike the standard LSH which only utilizes the linear distance calculation, in CFLSH, the short-listing step from a pool of pre-selected candidates filtered by hash functions to the final nearest neighbor set relies upon the frequency of collision along with distance information. We evaluate CFLSH versus the standard LSH using the L1 and cosine distances, for predicting acute hypotensive episodes on arterial blood pressure time series data extracted from the MIMIC II database. Our results show that CFLSH for the L1 distance has a higher prediction accuracy and further accelerates the sub-linear querying time obtained by the standard LSH.

Analysis of locality-sensitive hashing for fast critical event prediction on physiological time series.

We apply the sublinear time, scalable locality-sensitive hashing (LSH) and majority discrimination to the problem of predicting critical events based on physiological waveform time series. Compared to using the linear exhaustive k-nearest neighbor search, our proposed method vastly speeds up prediction time up to 25 times while sacrificing only 1% of accuracy when demonstrated on an arterial blood pressure dataset extracted from the MIMIC2 database. We compare two widely used variants of LSH, the bit sampling based (L1LSH) and the random projection based (E2LSH) methods to measure their direct impact on retrieval and prediction accuracy. We experimentally show that the more sophisticated E2LSH performs worse than L1LSH in terms of accuracy, correlation, and the ability to detect false negatives. We attribute this to E2LSH's simultaneous integration of all dimensions when hashing the data, which actually makes it more impotent against common noise sources such as data misalignment. We also demonstrate that the deterioration of accuracy due to approximation at the retrieval step of LSH has a diminishing impact on the prediction accuracy as the speed up gain accelerates.

WebUser: mining unexpected web usage

Web usage mining has been much concentrated on the discovery of relevant user behaviours from web access record data. In this paper, we present WebUser, an approach to discover unexpected usage in web access log. We present a belief-driven method for extracting unexpected web usage sequences, where the belief system consists of a temporal relation and semantics constrained sequence rules acquired with respect to prior knowledge. Our experiments show the effectiveness and usefulness of the proposed approach. Furthermore, discovered rules of unexpected web usage can be used for web content personalisation and recommendation, site structure optimisation and critical event prediction.

A model-based approach for systematic risk analysis

A modelling approach based on a mixed structural/functional graph that allows the simultaneous representation of the structural and the functional aspects of a system is presented. This model provides a framework to formalize risk analysis while being as simple as possible. The focus of the proposed approach is to achieve in a common analysis a functional failure analysis and the identification of physical damage, these being the two complementary parts of a full risk analysis. The resulting failure propagation graph is useful for model-based reasoning for fault diagnosis, the detection of dangerous situations, and the prediction of critical events. In addition, the approach is modular and reusable. The approach is illustrated by its use on an example.

Long-term outcome of patients with silent versus symptomatic ischemia six months after percutaneous coronary intervention and stenting

ObjectivesWe sought to evaluate the incidence of silent ischemia versus symptomatic ischemia six months after percutaneous coronary intervention (PCI) and its impact on prognosis and to test the utility of myocardial perfusion single-photon emission computed tomography (SPECT), or MPS, for risk stratification in these patients. BackgroundSilent ischemia is frequent after PCI. However, little is known about silent ischemia and long-term outcome after PCI and stenting. MethodsIn 356 consecutive patients with successful PCI and stenting and follow-up MPS after six months, long-term follow-up (4.1 ± 0.3 years) was performed. The MPS images were interpreted by defining summed stress, rest, and difference scores (summed difference score [SDS] = extent of ischemia) and related to symptoms and outcome. Critical events included cardiac death, myocardial infarction, and target vessel revascularization. ResultsEighty-one patients (23%) had evidence of target vessel ischemia, which was silent in 62%. The only independent predictor of silent ischemia was SDS (odds ratio 0.64, p = 0.001). During follow-up, 67 critical events occurred. For patients with an SDS of 0, 1–4, and >4, the critical event rates were 17%, 29%, and 69%, respectively. Similarly, patients without ischemia, silent ischemia, and symptomatic ischemia had 17%, 32%, and 52% of critical events, respectively. Diabetes (relative risk 1.98, p = 0.03) and SDS (relative risk 1.2, p < 0.001) were independent predictors of critical events. The MPS image added incremental information for the prediction of critical events. ConclusionsSix months after PCI and stenting, 23% of patients had target vessel ischemia, which was silent in 62%. Silent ischemia predicted a worse outcome than did no ischemia and tended to have a better outcome than symptomatic ischemia. This was closely related to the extent of ischemia. The SDS added incremental value to pre-scan findings with respect to diagnosis and prognosis, indicating the utility of MPS for risk stratification after PCI and stenting.