Time-varying Data Research Articles

Radiotherapy Standardisation and Artificial Intelligence within the National Cancer Institute's Clinical Trials Network

Artificial intelligence in healthcare refers to the use of complex algorithms designed to conduct certain tasks in an automated manner. Artificial intelligence has a transformative power in radiation oncology to improve the quality and efficiency of patient care, given the increase in volume and complexity of digital data, as well as the multi-faceted and highly technical nature of this field of medicine. However, artificial intelligence alone will not be able to fix healthcare's problem, because new technologies bring unexpected and potentially underappreciated obstacles. The inclusion of multicentre datasets, the incorporation of time-varying data, the assessment of missing data as well as informative censoring and the addition of clinical utility could significantly improve artificial intelligence models. Standardisation plays a crucial, supportive and leading role in artificial intelligence. Clinical trials are the most reliable method of demonstrating the efficacy and safety of a treatment or clinical approach, as well as providing high-level evidence to justify artificial intelligence. The National Surgical Adjuvant Breast and Bowel Project, the Radiation Therapy Oncology Group and the Gynecologic Oncology Group collaborated to form NRG Oncology (acronym NRG derived from the names of the parental groups). NRG Oncology is one of the adult cancer clinical trial groups containing radiotherapy specialty of the National Cancer Institute's Clinical Trials Network (NCTN). Standardisation from NRG/NCTN has the potential to reduce variation in clinical treatment and patient outcome by eliminating potential errors, enabling broader application of artificial intelligence tools. NCTN, NRG and Imaging and Radiation Oncology Core are in a unique position to help with standards development, advocacy and enforcement, all of which can benefit from artificial intelligence, as artificial intelligence has the ability to improve trial success rates by transforming crucial phases in clinical trial design, from study planning through to execution. Here we will examine: (i) how to conduct technical and clinical evaluations before adopting artificial intelligence technologies, (ii) how to obtain high-quality data for artificial intelligence, (iii) the NCTN infrastructure and standards, (iv) radiotherapy standardisation for clinical trials and (v) artificial intelligence applications in standardisation.

Visual Analysis of Traditional Chinese Medicine Health Records

<p indent="0mm">Traditional Chinese medicine is a profound source of Chinese culture, and studying health records is an effective means for traditional Chinese medicine inheritance and advancement. A visual analysis method is proposed for traditional Chinese medicine health records to analyze multivariate, multimodal, time-varying health record data, and studying medicines in high-dimensional symptom spaces. With multiple linked views composed by a flow chart, dimensionality reduction plots, and lab test plots, aided by brushing-and-linking interactions, the visual analysis method supports medical experts to practice the holistic view and the theory of syndrome differentiation in the analysis. A perception-inspired comparative visual mapping and interaction is designed to investigate the integration of traditional Chinese medicine and modern medicine. The analysis of three cases of various kidney diseases treated by a famous doctor demonstrates that proposed method is promising in traditional Chinese medicine inheritance, and mining core prescriptions to design new ones.

Characterization of far-field Mekong freshwater mass transport in the southern South China Sea using satellite gravimetry

Estuarine freshwater discharge to the ocean and its oceanic transport represent a dynamic interaction between freshwater and ambient seawater. While a near-field estuarine freshwater mass propagation within 350 km away from estuary has been investigated via in-situ data and modeling, its far-field propagation remains unexplored. This paper, for the first time, explores the potential usage of Mekong estuarine-discharged freshwater mass in the ocean inferred from time-variable satellite gravimetry data (e.g. Gravity Recovery and Climate Experiment (GRACE)) to characterize spread pattern and transport paths, and propagation direction and time duration in the entire Sunda Shelf and southern South China Sea. Through a cross-correlation analysis, spatial distribution of maximum cross-correlation coefficient and time lag between in-situ Mekong estuarine runoff and gravimetry-inferred estuarine-discharged freshwater mass on the ocean surface were calculated. It was found that the gravimetry-inferred time lags were largely consistent with the freshwater ages derived by isotope analysis sampled from the campaign in September 2007. During 2005–2013, the long-term spatial patterns of the maximum cross-correlation coefficient generated over sliding five four-year time spans yielded eastward clockwise and westward counterclockwise freshwater transport pathways, in summer and winter respectively, in the Sunda Shelf and the southern South China Sea. The long-term spatial patterns of the time lag showed less than a month was required for the freshwater mass transporting in the northeast direction, while the time lag lengthened progressively from a month to four months, clockwise from the northeast to the southwest direction. This could be explained by the monthly ocean surface current averaged between 2005 and 2013, simulated by the Estimating the Circulation and Climate of the Ocean (ECCO) model. Using eight one-year time spans between 2005 and 2013, the short-term spatial patterns of the maximum cross-correlation coefficient and the time lag varied considerably, which were likely attributable to global climate variability. As usual, the western boundary current in the southwestern South China Sea should be enhanced in La Nina year. However, in the strong La Niña event during 2010–2011, the weakened southwest current in the western Sunda Shelf enabled faster southward freshwater mass transport, resulting in shorter time lag (within three months) than that of the long-term one.

Multivariate batch to batch optimisation of fermentation processes to improve productivity

Increasing the productivity of batch processes presents a complex challenge, with difficulties resulting from, amongst other things, the time-varying and non-linear characteristics that such processes exhibit. This paper presents an innovative optimisation technique which utilises a data-driven Gaussian process regression model, built on time-varying and non-linear historical process data, to iteratively increase batch productivity from one cycle to the next. Specifically, productivity is increased by making appropriate adjustments to batch cycle time and the trajectory of a manipulated variable. The capabilities of the proposed method are demonstrated using two benchmark fermentation simulations, Penicillin production (Pensim) and Saccharomyces Cerevisiae, where it is shown to achieve increases in productivity of between 60% and 97% compared with what was achieved using ‘golden batch’ conditions.

Unsteady phenomena at the combustor-turbine interface

The combustor-turbine interface in a gas turbine is characterised by complex, highly unsteady flows. In a combined experimental and large eddy simulation (LES) study including realistic combustor geometry, the standard model of secondary flows in the nozzle guide vanes (NGV) is found to be oversimplified. A swirl core is created in the combustion chamber which convects into the first vane passages. Four main consequences of this are identified: variation in vane loading; unsteady heat transfer on vane surfaces; unsteadiness at the leading edge horseshoe vortex, and variation in the position of the passage vortex. These phenomena occur at relatively low frequencies, from 50–300 Hz. It seems likely that these unsteady phenomena result in non-optimal film cooling, and that by reducing unsteadiness designs with greater cooling efficiency could be achieved. Measurements were performed in a high speed test facility modelling a large industrial gas turbine with can combustors, including nozzle guide vanes and combustion chambers. Vane surfaces and endwalls of a nozzle guide vane were instrumented with 384 high speed thin film heat flux gauges, to measure unsteady heat transfer. The high resolution of measurements was such to allow direct visualisation in time of large scale turbulent structures over the endwalls and vane surfaces. A matching LES simulation was carried out in a domain matching experimental conditions including upstream swirl generators and transition duct. Data reduction allowed time-varying LES data to be recorded for several cycles of the unsteady phenomena observed. The combination of LES and experimental data allows physical explanation and visualisation of flow events.

State reach and development in Africa since the 1960s: new data and analysis

AbstractProminent arguments hold that African states’ geography limits state capacity, impedes public service provision, and slows economic development. To test this argument, I collect comprehensive panel data on a proxy of local state capacity, travel times to national and regional capitals. These are computed on a yearly 5 × 5 km grid using time-varying data on roads and administrative units (1966–2016). I use these data to estimate the effect of changes in travel times to capitals on local education provision, infant mortality rates, and nightlight emissions. Within the same location, decreases in travel times to its capitals are robustly associated with improved development outcomes. The article advances the measurement of state capacity and contributes to understanding its effects on human welfare.

STSRNet: Deep Joint Space-Time Super-Resolution for Vector Field Visualization.

We propose STSRNet, a joint space-time super-resolution deep learning based model for time-varying vector field data. Our method is designed to reconstruct high temporal resolution and high spatial resolution vector fields sequence from the corresponding low-resolution key frames. For large scale simulations, only data from a subset of time steps with reduced spatial resolution can be stored for post hoc analysis. In this article, we leverage a deep learning model to capture the nonlinear complex changes of vector field data with a two-stage architecture: the first stage deforms a pair of low spatial resolution (LSR) key frames forward and backward to generate the intermediate LSR frames, and the second stage performs spatial super-resolution to output the high-resolution sequence. Our method is scalable and can handle different datasets. We demonstrate the effectiveness of our framework with several datasets through quantitative and qualitative evaluations.

Monitoring calf circumference: changes during prolonged constrained sitting

Prolonged sitting has been associated with negative health effects; however, short-term time-varying exposure and response data is lacking. Twenty-two young and healthy participants were seated for 2 hours with the instruction to avoid the confounding effects of large leg movements while calf circumference, perceived discomfort, and lower limb muscle activity were collected. Calf circumference increased significantly (0.90 ± 0.32 cm) during sitting with no statistical differences between sexes. Perceived discomfort increased significantly over time in the low back and gluteal regions (p = 0.001–0.072, η p 2=0.080–0.360). On average, it took 20.31 ± 10.87 minutes of walking for calf measures to return to pre-sitting baseline. These results suggest that sitting for 2 hours without activity breaks may not be advisable and that recovery may take longer than expected. The exposure/response data from this study may be helpful in the design of future studies, with a larger and more general population, aiming to better define recommended duration/activity ratios for sitting-focused occupations. Practitioner summary: Leg swelling is a concern in prolonged sitting. In this study of young, healthy participants, we found a 2 hour constrained sitting exposure (controlling for large leg movements) induced significant increases in calf circumference that took an average of 20.31 ± 10.87 min of walking to return to baseline. Abbreviations: FMD: flow-mediated dilation, GSC: gastrocnemius; TA: tibialis anterior; EMG: electromyography; VAS: visual analog scale; MVC: maximum voluntary contractions.

3D structural vibration identification from dynamic point clouds

Video-based measurement has received increased attention for modal analysis and nondestructive evaluation, playing an important role in the development of the next-generation structural sensing technologies. As these techniques have evolved, more quantitative approaches based on computer vision techniques have emerged on full-field unsupervised structural identification, exploiting the benefits provided by the use of video cameras such as high spatial sensor density and low installation costs. More recent work has started to explore the use of laser point cloud data for 3D mapping of scenes and structures. Sensors such as LIDAR provide huge amounts of measurements at high spatial resolution from which it is possible to estimate accurate structural geometry for applications such as the generation of CAD models. Unfortunately to-date, the frame rate and depth resolution of LIDAR and other sensors capable of 3D geometry measurements has not been sufficient for measuring structural dynamics. In this paper, we introduce an approach for efficient and extremely high resolution 3D structural dynamic identification/modal analysis from point cloud data acquired using a commercial, low-cost, time-of-flight imager. Vibration mode shapes and modal coordinates are extracted from this data by creating virtual Lagrangian sensors based on the point clouds parameters. First, time-varying point cloud data are collected from a vibrating structure. Then, a mesh of virtual sensors is created based on the dynamic point cloud data for tracking the structure’s displacement over time. Next solutions to the blind source separation problem are employed to estimate high resolution 3D mode shapes, modal coordinates, and resonant frequencies. We demonstrate the potential of our proposed approach on laboratory tests and compare the results to the data collected from conventional laser displacement sensors. This technique represents an advance towards efficiently exploring the full advantages of using dynamic point cloud data for practical monitoring applications and has the potential to be extended for a wide range of 3D motion decomposition problems.

When Internet of Things Meets e-Health: An Indoor Temperature Monitoring and Control Approach

Ambient temperature is closely related to human health. Maintaining a comfortable and stable temperature is crucial for the treatment, rehabilitation and daily e-health care of patients, where the Internet of Things has been widely applied for this purpose. However, the existing research always separates the network design from temperature monitoring and control. In order to provide better e-health services, we design a solution to minimize the cost of energy consumption and delay. This work discusses a cyber-physical design for an indoor temperature monitoring system using a wireless sensor network. All sensors dynamically adjust the sleep/wake duty cycles and select the optimal anycast routing according to the sensed temperature. In addition, IoT-enabled heating, ventilation, and air conditioner (HVAC) systems for indoor temperature control have attracted unprecedented attention. HVAC can be connected to the Internet for weather and time-varying electricity price data. This work discusses the problem of minimizing the total energy cost of the HVAC system while keeping the indoor temperature within a prefixed range. The key idea is to leverage the time-varying electricity prices and do precooling/preheating using HVAC. The simulation results show that our temperature monitoring and control algorithm outperforms other heuristic schemes.

Interactive Visualization of Hyperspectral Images Based on Neural Networks.

It is challenging to interpret hyperspectral images in an intuitive and meaningful way, as they usually contain hundreds of dimensions. We develop a visualization tool for hyperspectral images based on neural networks, which allows a user to specify the regions of interest, select bands of interest, and obtain hyperspectral classification results in a scatterplot generated from hyperspectral features. A cascade neural network is trained to generate a scatterplot that matches the cluster centers labeled by the user. The inferred scatterplot not only shows the clusters of points, but also reveals relationships of substances. The trained neural network can be reused for time-varying hyperspectral data analysis without retraining. Our visualization solution can keep domain experts in the analytical loop and provide an intuitive analysis of hyperspectral images while identifying different substances, which are difficult to be realized using existing hyperspectral image analysis techniques.

Virtual Reality Visualization of CFD Simulated Blood Flow in Cerebral Aneurysms Treated with Flow Diverter Stents

Virtual reality (VR) has the potential to be a powerful tool for the visualization of simulated blood flow in cerebral aneurysms. This paper presents our study aimed at developing the VR visualization of computational fluid dynamics (CFD) simulations of cerebral aneurysms treated with flow-diverting (FD) stents. First, a spherical sidewall aneurysm located at a simplified internal carotid artery was considered for investigating the impact of stent deployment and positioning on the corresponding spatially time-varying blood flow behavior. The three-dimensional unsteady blood flow over a cardiac cycle was simulated numerically using the finite volume method, and the distributions of hemodynamic parameters inside the aneurysm sac, and on its wall, were presented with and without stent cases. Two stent positions, with and without a gap between the artery wall and stent, were considered to show the influence of correct stent position on aneurysm treatment. Second, a straightforward workflow was developed to import, process, and visualize the CFD analysis data in a VR environment by using open-source software with a high resolution. The Unity3D engine was used for displaying the processed animations in a VR environment operated on a head-mounted display (HMD). The refining process of each frame of time-varying CFD data was automated. The animated flow elements rendered in the VR environment were velocity vectors, velocity contours, streamlines, particle traces, and point clouds. CFD results showed that proper stenting facilitates thrombosis and occlusion of the aneurysm by modification of the flow patterns, which leads to lower inflow jet velocities into the aneurysm, longer turnover time, lower aneurysm-averaged kinetic energy, and lower wall shear stress. Additionally, the results indicated that a gap between the stent and the parent artery may lead to undesirable hemodynamic alterations. The VR visualization illustrated that the recognition of the potential in danger regions of aneurysms and the evaluation of the performance of FD stents in aneurysm treatment can be conducted without the need for several slices through the parent artery and aneurysm, as is required for traditional postprocessing methods. Through VR visualization, the details of the simulation results become readily available by navigating in the 3D animated flow elements using a high-degree-of-freedom headset.

Modeling and prediction of pressure injury in hospitalized patients using artificial intelligence

BackgroundHospital-acquired pressure injuries (PIs) induce significant patient suffering, inflate healthcare costs, and increase clinical co-morbidities. PIs are mostly due to bed-immobility, sensory impairment, bed positioning, and length of hospital stay. In this study, we use electronic health records and administrative data to examine the contributing factors to PI development using artificial intelligence (AI).MethodsWe used advanced data science techniques to first preprocess the data and then train machine learning classifiers to predict the probability of developing PIs. The AI training was based on large, incongruent, incomplete, heterogeneous, and time-varying data of hospitalized patients. Both model-based statistical methods and model-free AI strategies were used to forecast PI outcomes and determine the salient features that are highly predictive of the outcomes.ResultsOur findings reveal that PI prediction by model-free techniques outperform model-based forecasts. The performance of all AI methods is improved by rebalancing the training data and by including the Braden in the model learning phase. Compared to neural networks and linear modeling, with and without rebalancing or using Braden scores, Random forest consistently generated the optimal PI forecasts.ConclusionsAI techniques show promise to automatically identify patients at risk for hospital acquired PIs in different surgical services. Our PI prediction model provide a first generation of AI guidance to prescreen patients at risk for developing PIs.Clinical impactThis study provides a foundation for designing, implementing, and assessing novel interventions addressing specific healthcare needs. Specifically, this approach allows examining the impact of various dynamic, personalized, and clinical-environment effects on PI prevention for hospital patients receiving care from various surgical services.

Effects of transmission shaft flexibility on rolling element bearing tribodynamics in a high-performance transmission

In a high-performance transmission, rolling bearings play a crucial role in supporting the loaded shafts and gear mesh whilst minimising friction. Prediction of critical speeds and dynamic responses is important to avoid prolonged periods of elevated vibration amplitudes, which can contribute to premature bearing and gearbox failure. Nonlinearities such as backlash and time-varying stiffnesses can lead to complex coupled dynamics between gears and bearings. In the current study, an 8-speed transmission is discretized and modelled by the finite element method. The eigenproblem is solved using damped modal analysis considering gyroscopic effects. Time-varying stiffness data and backlashes are introduced as nonlinearities. The system response is compared for cases of linear and nonlinear bearings, and rigid and flexible shafts to ascertain the significance of elasticity. The flexible model improves the accuracy of bearing load oscillation by altering the transfer path between gears and bearings. This provides an advancement upon existing works, where shaft dynamics are often simplified. Furthermore, it is shown that the combination of local nonlinear effects and flexible components significantly changes the response of the system and the bearing life prediction for a system-level analysis.

Using machine learning to predict severe hypoglycaemia in hospital.

To predict the risk of hypoglycaemia using machine-learning techniques in hospitalized patients. We conducted a retrospective cohort study of patients hospitalized under general internal medicine (GIM) and cardiovascular surgery (CV) at a tertiary care teaching hospital in Toronto, Ontario. Three models were generated using supervised machine learning: least absolute shrinkage and selection operator (LASSO) logistic regression; gradient-boosted trees; and a recurrent neural network. Each model included baseline patient data and time-varying data. Natural-language processing was used to incorporate text data from physician and nursing notes. We included 8492 GIM admissions and 8044 CV admissions. Hypoglycaemia occurred in 16% of GIM admissions and 13% of CV admissions. The area under the curve for the models in the held-out validation set was approximately 0.80 on the GIM ward and 0.82 on the CV ward. When the threshold for hypoglycaemia was lowered to 2.9 mmol/L (52 mg/dL), similar results were observed. Among the patients at the highest decile of risk, the positive predictive value was approximately 50% and the sensitivity was 99%. Machine-learning approaches can accurately identify patients at high risk of hypoglycaemia in hospital. Future work will involve evaluating whether implementing this model with targeted clinical interventions can improve clinical outcomes.

Modeling Time-Varying Random Objects and Dynamic Networks

Samples of dynamic or time-varying networks and other random object data such as time-varying probability distributions are increasingly encountered in modern data analysis. Common methods for time-varying data such as functional data analysis are infeasible when observations are time courses of networks or other complex non-Euclidean random objects that are elements of general metric spaces. In such spaces, only pairwise distances between the data objects are available and a strong limitation is that one cannot carry out arithmetic operations due to the lack of an algebraic structure. We combat this complexity by a generalized notion of mean trajectory taking values in the object space. For this, we adopt pointwise Fréchet means and then construct pointwise distance trajectories between the individual time courses and the estimated Fréchet mean trajectory, thus representing the time-varying objects and networks by functional data. Functional principal component analysis of these distance trajectories can reveal interesting features of dynamic networks and object time courses and is useful for downstream analysis. Our approach also makes it possible to study the empirical dynamics of time-varying objects, including dynamic regression to the mean or explosive behavior over time. We demonstrate desirable asymptotic properties of sample based estimators for suitable population targets under mild assumptions. The utility of the proposed methodology is illustrated with dynamic networks, time-varying distribution data and longitudinal growth data.

Research on Adaptive Power Transmission Line Fault Inspection

Power transmission line is one of the most important infrastructures of power system, and its safety monitoring is of great significance. The conventional way of fault monitoring of power transmission lines by only setting threshold value on single temperature data of strain clamp turned out to multiple misjudgements and delayed alarms, causing the increment of operation risk of power transmission line. In this paper, various types of time-varying sensors data such as strain clamp temperature data, environmental data, and cable ampacity data are accounted. Also, an unsupervised machine learning algorithm - K-means clustering algorithm was introduced to build a discriminant model in detecting the defects of power transmission line. Experimental results proved that the proposed method is able to avoid delayed alarms as well as misjudgement incurred from conventional method. As a result, the operation safety of power transmission line and inspection efficiency can be improved. The inspection cost would be reduced as well.

Digital Twin-Based Safety Risk Coupling of Prefabricated Building Hoisting.

Safety management in hoisting is the key issue to determine the development of prefabricated building construction. However, the security management in the hoisting stage lacks a truly effective method of information physical fusion, and the safety risk analysis of hoisting does not consider the interaction of risk factors. In this paper, a hoisting safety risk management framework based on digital twin (DT) is presented. The digital twin hoisting safety risk coupling model is built. The proposed model integrates the Internet of Things (IoT), Building Information Modeling (BIM), and a security risk analysis method combining the Apriori algorithm and complex network. The real-time perception and virtual–real interaction of multi-source information in the hoisting process are realized, the association rules and coupling relationship among hoisting safety risk factors are mined, and the time-varying data information is visualized. Demonstration in the construction of a large-scale prefabricated building shows that with the proposed framework, it is possible to complete the information fusion between the hoisting site and the virtual model and realize the visual management. The correlative relationship among hoisting construction safety risk factors is analyzed, and the key control factors are found. Moreover, the efficiency of information integration and sharing is improved, the gap of coupling analysis of security risk factors is filled, and effective security management and decision-making are achieved with the proposed approach.

Experimental investigation of structural system capacity with multiple fatigue cracks

Few experimental data sets exist in the literature to support the development and evaluation of digital twins predicting structural degradation. The literature is especially sparse for system tests where multiple failures occur and interact. In this work, a laboratory-level experiment is conducted to mimic many of the properties of larger and more complex marine structures with redundant load paths, failure interaction, and component-to-system level integration. In the experiment, such properties are reflected by a hexagon tension specimen with four propagating fatigue cracks tested under displacement-controlled loading. The applied loading cycles and corresponding crack lengths are recorded as the major time-varying data of degradation, with the resisting force at maximum extension used as the system capacity. A novel computer vision method is used to measure the crack length. Strain gauges are also used to monitor the structure’s status. The experimental data is presented and analyzed in this paper. The resulting data sets can be used to evaluate the performance of different digital twin updating approaches.

Online domain description of big data based on hyperellipsoid models

Big data is usually massive, diverse, time-varying, and high-dimensional. The focus of this paper is on the domain description of big data, which is the basis for solving the above problems. This paper has three main contributions. Firstly, one hyperellipsoid model is proposed to analyze domain description of big data. The parameters of the hyperellipsoid model can be adaptively adjusted according to the proposed objective function without relying on manual parameter selection, which expands the application range of the model. Secondly, an improved FDPC algorithm is proposed to generate multiple hyperellipsoid models to approximate the spatial distribution of big data, thus improving the accuracy of domain description. Multiple hyperellipsoid models can not only greatly eliminate the spatial redundancy of the domain description based on one hyperellipsoid model, but also provide a feasible method for describing complex spatial distribution. Thirdly, an online domain description algorithm based on hyperellipsoid models is proposed, which improves the robustness of hyperellipsoid models on time-varying data. The parallel processing flow of the algorithm is given. In the experiment, synthetic instances and real-world datasets were applied to test the performance of hyperellipsoid models. By comparing LOF, OneClassSVM, SVDD and isolation forest, the performance of the proposed method is competitive and promising.