Search

This study aimed to investigate the application of label propagation techniques to propagate labels among photoplethysmogram (PPG) signals, particularly in imbalanced class scenarios and limited data availability scenarios, where clean PPG samples are significantly outnumbered by artifact-contaminated samples. We investigated a dataset comprising PPG recordings from 1571 patients, wherein approximately 82% of the samples were identified as clean, while the remaining 18% were contaminated by artifacts. Our research compares the performance of supervised classifiers, such as conventional classifiers and neural networks (Multi-Layer Perceptron (MLP), Transformers, Fully Convolutional Network (FCN)), with the semi-supervised Label Propagation (LP) algorithm for artifact classification in PPG signals. The results indicate that the LP algorithm achieves a precision of 91%, a recall of 90%, and an F1 score of 90% for the “artifacts” class, showcasing its effectiveness in annotating a medical dataset, even in cases where clean samples are rare. Although the K-Nearest Neighbors (KNN) supervised model demonstrated good results with a precision of 89%, a recall of 95%, and an F1 score of 92%, the semi-supervised algorithm excels in artifact detection. In the case of imbalanced and limited pediatric intensive care environment data, the semi-supervised LP algorithm is promising for artifact detection in PPG signals. The results of this study are important for improving the accuracy of PPG-based health monitoring, particularly in situations in which motion artifacts pose challenges to data interpretation.

BackgroundClinical decision-making is a complex cognitive process that relies on the interpretation of a large variety of data from different sources and involves the use of knowledge bases and scientific recommendations. The representation of clinical data plays a key role in the speed and efficiency of its interpretation. In addition, the increasing use of clinical decision support systems (CDSSs) provides assistance to clinicians in their practice, allowing them to improve patient outcomes. In the pediatric intensive care unit (PICU), clinicians must process high volumes of data and deal with ever-growing workloads. As they use multiple systems daily to assess patients’ status and to adjust the health care plan, including electronic health records (EHR), clinical systems (eg, laboratory, imaging and pharmacy), and connected devices (eg, bedside monitors, mechanical ventilators, intravenous pumps, and syringes), clinicians rely mostly on their judgment and ability to trace relevant data for decision-making. In these circumstances, the lack of optimal data structure and adapted visual representation hinder clinician’s cognitive processes and clinical decision-making skills. ObjectiveIn this study, we designed a prototype to optimize the representation of clinical data collected from existing sources (eg, EHR, clinical systems, and devices) via a structure that supports the integration of a home-developed CDSS in the PICU. This study was based on analyzing end user needs and their clinical workflow. MethodsFirst, we observed clinical activities in a PICU to secure a better understanding of the workflow in terms of staff tasks and their use of EHR on a typical work shift. Second, we conducted interviews with 11 clinicians from different staff categories (eg, intensivists, fellows, nurses, and nurse practitioners) to compile their needs for decision support. Third, we structured the data to design a prototype that illustrates the proposed representation. We used a brain injury care scenario to validate the relevance of integrated data and the utility of main functionalities in a clinical context. Fourth, we held design meetings with 5 clinicians to present, revise, and adapt the prototype to meet their needs. ResultsWe created a structure with 3 levels of abstraction—unit level, patient level, and system level—to optimize clinical data representation and display for efficient patient assessment and to provide a flexible platform to host the internally developed CDSS. Subsequently, we designed a preliminary prototype based on this structure. ConclusionsThe data representation structure allows prioritizing patients via criticality indicators, assessing their conditions using a personalized dashboard, and monitoring their courses based on the evolution of clinical values. Further research is required to define and model the concepts of criticality, problem recognition, and evolution. Furthermore, feasibility tests will be conducted to ensure user satisfaction.

When dealing with clinical text classification on a small dataset, recent studies have confirmed that a well-tuned multilayer perceptron outperforms other generative classifiers, including deep learning ones. To increase the performance of the neural network classifier, feature selection for the learning representation can effectively be used. However, most feature selection methods only estimate the degree of linear dependency between variables and select the best features based on univariate statistical tests. Furthermore, the sparsity of the feature space involved in the learning representation is ignored. Goal: Our aim is, therefore, to access an alternative approach to tackle the sparsity by compressing the clinical representation feature space, where limited French clinical notes can also be dealt with effectively. Methods: This study proposed an autoencoder learning algorithm to take advantage of sparsity reduction in clinical note representation. The motivation was to determine how to compress sparse, high-dimensional data by reducing the dimension of the clinical note representation feature space. The classification performance of the classifiers was then evaluated in the trained and compressed feature space. Results: The proposed approach provided overall performance gains of up to 3% for each test set evaluation. Finally, the classifier achieved 92% accuracy, 91% recall, 91% precision, and 91% f1-score in detecting the patient’s condition. Furthermore, the compression working mechanism and the autoencoder prediction process were demonstrated by applying the theoretic information bottleneck framework. Clinical and Translational Impact Statement— An autoencoder learning algorithm effectively tackles the problem of sparsity in the representation feature space from a small clinical narrative dataset. Significantly, it can learn the best representation of the training data because of its lossless compression capacity compared to other approaches. Consequently, its downstream classification ability can be significantly improved, which cannot be done using deep learning models.

IntroductionLow cardiac output syndrome in the postoperative period after cardiac surgery leads to an increase in tissue oxygen extraction, assessed by the oxygen extraction ratio. Measurement of the oxygen extraction ratio requires blood gases to be taken. However, the temperature of the skin and various parts of the body is a direct result of blood flow distribution and can be monitored using infrared thermography. Thus, we conducted a prospective clinical study to evaluate the correlation between the thermal gradient obtained by infrared thermography and the oxygen extraction ratio in children at risk for low cardiac output after cardiac surgery.MethodsChildren aged 0 to 18 years, having undergone cardiac surgery with cardio-pulmonary bypass in a pediatric intensive care unit were included in the study. One to 4 thermal photos were taken per patient using the FLIR One Pro thermal imaging camera. The thermal gradient between the central temperature of the inner canthus of the eye and the peripheral temperature was compared to the concomitant oxygen extraction ratio calculated from blood gases.Results41 patients were included with a median age of 6 months (IQR 3–48) with median Risk Adjustment for Congenital Heart Surgery-1 score was 2 (IQR 2–3). Eighty nine thermal photos were analyzed. The median thermal gradient was 2.5 °C (IQR 1,01–4.04). The median oxygen extraction ratio was 35% (IQR 26–42). Nine patients had an oxygen extraction ratio ≥ 50%. A significant but weak correlation was found between the thermal gradient and the oxygen extraction ratio (Spearman's test p = 0.25, p = 0.016). Thermal gradient was not correlated with any other clinical or biologic markers of low cardiac output. Only young age was an independent factor associated with an increase in the thermal gradient.ConclusionIn this pilot study, which included mainly children without severe cardiac output decrease, a significant but weak correlation between thermal gradient by infrared thermography and oxygen extraction ratio after pediatric cardiac surgery was observed. Infrared thermography is a promising non-invasive technology that could be included in multimodal monitoring of postoperative cardiac surgery patients. However, a clinical trial including more severe children is needed.

The rapid progress in clinical data management systems and artificial intelligence approaches enable the era of personalized medicine. Intensive care units (ICUs) are ideal clinical research environments for such development because they collect many clinical data and are highly computerized. Goal: We designed a retrospective clinical study on a prospective ICU database using clinical natural language to help in the early diagnosis of heart failure in critically ill children. Methods: The methodology consisted of empirical experiments of a learning algorithm to learn the hidden interpretation and presentation of the French clinical note data. This study included 1386 patients' clinical notes with 5444 single lines of notes. There were 1941 positive cases (36% of total) and 3503 negative cases classified by two independent physicians using a standardized approach. Results: The multilayer perceptron neural network outperforms other discriminative and generative classifiers. Consequently, the proposed framework yields an overall classification performance with 89% accuracy, 88% recall, and 89% precision. Conclusions: This study successfully applied learning representation and machine learning algorithms to detect heart failure in a single French institution from clinical natural language. Further work is needed to use the same methodology in other languages and institutions.

The temperature distribution at the skin surface could be a useful tool to monitor changes in cardiac output. Goal: The aim of this study was to explore infrared thermography as a method to analyze temperature profiles of critically ill children. Methods: Patients admitted to the pediatric intensive care unit (PICU) were included in this study. An infrared sensor was used to take images in clinical conditions. The infrared core and limb temperatures (θc & θl) were extracted, as well as temperatures along a line drawn between these two regions. Results: The median [interquartile range] θc extracted from the images was 33.88°C [32.74-34.19] and the median θl was 30.21°C [28.89-33.13]. There was a good correlation between the θc and the clinical axillary temperature (rho = 0.39, p-value = 0.016). There was also a good correlation between the θc and θl (rho = 0.66, p-value = 1.2 e−05). Conclusion: Thermography was found to be effective to estimate the body temperature. Correlation with specific clinical conditions needs further study.

In children, vital distress events, particularly respiratory, go unrecognized. To develop a standard model for automated assessment of vital distress in children, we aimed to construct a prospective high-quality video database for critically ill children in a pediatric intensive care unit (PICU) setting. The videos were acquired automatically through a secure web application with an application programming interface (API). The purpose of this article is to describe the data acquisition process from each PICU room to the research electronic database. Using an Azure Kinect DK and a Flir Lepton 3.5 LWIR attached to a Jetson Xavier NX board and the network architecture of our PICU, we have implemented an ongoing high-fidelity prospectively collected video database for research, monitoring, and diagnostic purposes. This infrastructure offers the opportunity to develop algorithms (including computational models) to quantify vital distress in order to evaluate vital distress events. More than 290 RGB, thermographic, and point cloud videos of each 30 s have been recorded in the database. Each recording is linked to the patient’s numerical phenotype, i.e., the electronic medical health record and high-resolution medical database of our research center. The ultimate goal is to develop and validate algorithms to detect vital distress in real time, both for inpatient care and outpatient management.

Falls are a growing issue in society and has become a hot topic in the healthcare domain. Falls are more likely to occur to due to age or health problems such as cardiovascular issues and muscle weakness. In this work we focus on fall detection. The aftereffects of falls often lead to the use of prescription pain medications. We are motivated to help prevent suicide attempts by overdose in the Canadian correctional services. Most previous studies were based on hand-crafted features which limit the robustness and generality of the system. We therefore propose a general vision-based system, using Spatial Temporal Graph Convolutional Networks (ST-GCN). This system has proven its efficiency and robustness in the action recognition domain. Contrary to previous works, this model can be applied directly to new data without the need to retrain the model while offering good accuracy. Additionally, with the help of transfer learning we can solve the insufficient data problem. By using three public datasets: the NTU RGB-D dataset, the TST Fall detection dataset v2 and the Fallfree dataset to validate our method, we achieved a 100% accuracy, surpassing the state-of-the-art.

For skeleton-based action recognition from depth cameras, distinguishing object-related actions with similar motions is a difficult task. The other available video streams (RGB, infrared, depth) may provide additional clues, given an appropriate feature fusion strategy. We propose a modular network combining skeleton and infrared data. A pre-trained 2D convolutional neural network (CNN) is used as a pose module to extract features from skeleton data. A pre-trained 3D CNN is used as an infrared module to extract visual features from videos. Both feature vectors are then fused and exploited jointly using a multilayer perceptron (MLP). The 2D skeleton coordinates are used to crop a region of interest around the subjects for the infrared videos. Infrared is favored over RGB, as it is less affected by illumination conditions and usable in the dark. We are the first to combine infrared and skeleton data. We evaluate our method on the NTU RGB+D dataset, the largest dataset for human action recognition from depth cameras. We perform extensive ablation studies. In particular, we show the strong contributions of our cropping strategy and pre-training on action classification accuracy. We also test various feature fusion schemes. Feature sum on an element-wise level yields the best results. Our method achieves state-of-the-art performances on NTU RBG+D.

In this paper, we propose a solution for video transmission over a low-bandwidth network that enables a physician to take in charge of remote trauma patient. We propose and analyze a method based on an H.264 compression scheme that relies on transmitting high-quality video of a moving and dynamic region of interest while scarifying quality in the background. Our method is motivated by the problem of limited bandwidth usually encountered in air-to-ground communication channels. We propose to use a region of interest with smoothed edges to increase the video quality of the transition between the regions of various qualities. The moving region of interest, covering the torso and the head, is segmented and tracked by using the skeleton information provided by a Kinect camera. Our proposed compression scheme respects the real-time, low-complexity, and interoperability constraints. We have analyzed the results of our method obtained with various bit rate targets and have shown that a visual assessment of a patient is achievable over very low bandwidth.

In a situation of respiratory failure (RF), patients show signs of increased work of breathing leading to the involvement of accessory respiratory muscles and a desynchronization between rib cage and abdomen known as thoraco-abdominal asynchrony (TAA). Proper assessment of these signs requires sufficiently skilled and trained medical staff. However, human assessment is subjective and is practically impossible to audit. A new non-contact method is proposed for TAA visualization and quantification, in children with RF. The surface variations are analyzed by calculating the 3-dimensional motion of the thorax and abdomen regions during the breathing process. A high-fidelity mannequin was used to simulate thoraco-abdominal asynchrony. The proposed system uses depth information recorded by an RGB-D (Red Green Blue-Depth) camera. Furthermore, surface displacement was calculated in four simulated modes from the normal to the severe TAA mode. Respiratory rates were also calculated based on the analysis of the surface movements. The proposed method was compared to a highly precise laser-ranging system with 1 mm resolution. The resulting root mean square deviation (RMSD) showed an error of 1.78 ml in normal mode, 2.83 mm in mild mode, 2.23 mm in severe mode and 2.34 mm in irregular mode. The results showed a high correlation between the two methods in estimating the retraction distance and respiratory rate (ρ >0.985).

BackgroundIn an intensive care units, experts in mechanical ventilation are not continuously at patient's bedside to adjust ventilation settings and to analyze the impact of these adjustments on gas exchange. The development of clinical decision support systems analyzing patients' data in real time offers an opportunity to fill this gap.ObjectiveThe objective of this study was to determine whether a machine learning predictive model could be trained on a set of clinical data and used to predict transcutaneous hemoglobin oxygen saturation 5 min (5min SpO2) after a ventilator setting change.Data sourcesData of mechanically ventilated children admitted between May 2015 and April 2017 were included and extracted from a high-resolution research database. More than 776,727 data rows were obtained from 610 patients, discretized into 3 class labels (< 84%, 85% to 91% and c92% to 100%).Performance metrics of predictive modelsDue to data imbalance, four different data balancing processes were applied. Then, two machine learning models (artificial neural network and Bootstrap aggregation of complex decision trees) were trained and tested on these four different balanced datasets. The best model predicted SpO2 with area under the curves < 0.75.ConclusionThis single center pilot study using machine learning predictive model resulted in an algorithm with poor accuracy. The comparison of machine learning models showed that bagged complex trees was a promising approach. However, there is a need to improve these models before incorporating them into a clinical decision support systems. One potentially solution for improving predictive model, would be to increase the amount of data available to limit over-fitting that is potentially one of the cause for poor classification performances for 2 of the three class labels.

Assessment of respiratory function allows early detection of potential disorders in the respiratory system and provides useful information for medical management. There is a wide range of applications for breathing assessment, from measurement systems in a clinical environment to applications involving athletes. Many studies on pulmonary function testing systems and breath monitoring have been conducted over the past few decades, and their results have the potential to broadly impact clinical practice. However, most of these works require physical contact with the patient to produce accurate and reliable measures of the respiratory function. There is still a significant shortcoming of non-contact measuring systems in their ability to fit into the clinical environment. The purpose of this paper is to provide a review of the current advances and systems in respiratory function assessment, particularly camera-based systems. A classification of the applicable research works is presented according to their techniques and recorded/quantified respiration parameters. In addition, the current solutions are discussed with regards to their direct applicability in different settings, such as clinical or home settings, highlighting their specific strengths and limitations in the different environments.

After three decades from the public debut of cellular networks, there are hardly parts of populated lands where cellular coverage is absent. Day after day, mobile users have been provided with wider range of services at higher speed. Today, long terminal evolution (LTE) networks support broadband connectivity for users moving as fast as 350km/h, and the support for speeds up to 500km/h is under consideration. Unfortunately, none of these efforts were aimed at airborne travelers due to the lack of aerial coverage. Provided that 5G networks are meant to provide anywhere and anytime connectivity for anyone, many operators are providing the free onboard Wi-Fi through proprietary terrestrial networks or satellite links. Unfortunately, both of these solutions have serious drawbacks, where the latter provides very limited speed, and the former is expensive and unscalable. In this paper, we discuss the technical possibilities of enhancing the existing LTE infrastructure for air to ground communications. We identify the major challenges and obstacles in this path, such as uplink/downlink interferences, frequent roaming, large Doppler effect, and channel degradation. We also discuss appropriate solutions to counteract them using some of the emerging antenna, signal processing, beamforming, and multi-beaming ideas.

In the domain of human action recognition, existing works mainly focus on using RGB, depth, skeleton and infrared data for analysis. While these methods have the benefit of being non-invasive, they can only be used within limited setups, are prone to issues such as occlusion and often need substantial computational resources. In this work, we address human action recognition through inertial sensor signals, which have a vast quantity of practical applications in fields such as sports analysis and human-machine interfaces. For that purpose, we propose a new learning framework built around a 1D-CNN architecture, which we validated by achieving very competitive results on the publicly available UTD-MHAD dataset. Moreover, the proposed method provides some answers to two of the greatest challenges currently faced by action recognition algorithms, which are (1) the recognition of high-level activities and (2) the reduction of their computational cost in order to make them accessible to embedded devices. Finally, this paper also investigates the tractability of the features throughout the proposed framework, both in time and duration, as we believe it could play an important role in future works in order to make the solution more intelligible, hardware-friendly and accurate.

Higher plants are an integral part of strategies for sustained human presence in space. Space-based greenhouses have the potential to provide closed-loop recycling of oxygen, water and food. Plant monitoring systems with the capacity to remotely observe the condition of crops in real-time within these systems would permit operators to take immediate action to ensure optimum system yield and reliability. One such plant health monitoring technique involves the use of reporter genes driving fluorescent proteins as biological sensors of plant stress. In 2006 an initial prototype green fluorescent protein imager system was deployed at the Arthur Clarke Mars Greenhouse located in the Canadian High Arctic. This prototype demonstrated the advantageous of this biosensor technology and underscored the challenges in collecting and managing telemetric data from exigent environments. We present here the design and deployment of a second prototype imaging system deployed within and connected to the infrastructure of the Arthur Clarke Mars Greenhouse. This is the first imager to run autonomously for one year in the un-crewed greenhouse with command and control conducted through the greenhouse satellite control system. Images were saved locally in high resolution and sent telemetrically in low resolution. Imager hardware is described, including the custom designed LED growth light and fluorescent excitation light boards, filters, data acquisition and control system, and basic sensing and environmental control. Several critical lessons learned related to the hardware of small plant growth payloads are also elaborated.