Design of Hardware Accelerators for Optimized and Quantized Neural Networks to Detect Atrial Fibrillation in Patch ECG Device with RISC-V.

Atrial Fibrillation (AF) is one of the most common heart arrhythmias. It is known to cause up to 15% of all strokes. In current times, modern detection systems for arrhythmias, such as single-use patch electrocardiogram (ECG) devices, have to be energy efficient, small, and affordable. In this work, specialized hardware accelerators were developed. First, an artificial neural network (NN) for the detection of AF was optimized. Special attention was paid to the minimum requirements for the inference on a RISC-V-based microcontroller. Hence, a 32-bit floating-point-based NN was analyzed. To reduce the silicon area needed, the NN was quantized to an 8-bit fixed-point datatype (Q7). Based on this datatype, specialized accelerators were developed. Those accelerators included single-instruction multiple-data (SIMD) hardware as well as accelerators for activation functions such as sigmoid and hyperbolic tangents. To accelerate activation functions that require the e-function as part of their computation (e.g., softmax), an e-function accelerator was implemented in the hardware. To compensate for the losses of quantization, the network was expanded and optimized for run-time and memory requirements. The resulting NN has a 7.5% lower run-time in clock cycles (cc) without the accelerators and 2.2 percentage points (pp) lower accuracy compared to a floating-point-based net, while requiring 65% less memory. With the specialized accelerators, the inference run-time was lowered by 87.2% while the F1-Score decreased by 6.1 pp. Implementing the Q7 accelerators instead of the floating-point unit (FPU), the silicon area needed for the microcontroller in 180 nm-technology is below 1 mm2.

Application of internal electrodes to the oesophageal and tracheal tube in an animal trial: evaluation of its clinical and technical potentiality in electrical impedance tomography.

Electrical impedance tomography (EIT) is of potential medical interest e.g., to optimize ventilator settings during mechanical ventilation. Nevertheless there are still several challenges. Although electrode belts are commonly used and promoted, they are not necessarily adequate for the long-term monitoring of patients in intensive-care units (ICU). ICU patients are usually equipped with breathing tubes and feeding tubes, ideal surfaces to attach EIT electrodes to. The aim of our study was therefore to examine the potentiality of internal electrodes in a porcine animal trial. Following an animal trial protocol studying acute lung injury, additional EIT measurements were obtained both with conventional electrodes set upon a rubber belt and after having moved the electrodes internally in seven pigs. For this reason the two most dorsally located electrodes were selected. An adjacent stimulation and measurement pattern was used, and resulting voltages in the time and frequency domains as well as within reconstructed images were examined to compare perfusion and ventilation data qualitatively and quantitatively. Particularly, lung morphology as well as signal strength for both the mediastinal and lung region were studied. All animals were submitted to the additional protocol without any adverse events. Distinguishability of lungs was improved in reconstructed frames. The resulting sensitivity of measured electrical impedance was enhanced around the mediastinal region and even cardiac-related activity was significantly increased by a factor of up to 6. In conclusion the application of internal electrodes appears to be beneficial for diverse clinical purposes and should be addressed in further studies.

Methods and evaluation of physiological measurements with acoustic stimuli-a systematic review.

Objective. The detection of psychological loads, such as stress reactions, is receiving greater attention and social interest, as stress can have long-term effects on health O'Connor, Thayer and Vedhara (2021Ann. Rev. Psychol.72, 663-688). Acoustic stimuli, especially noise, are investigated as triggering factors. The application of physiological measurements in the detection of psychological loads enables the recording of a further quantitative dimension that goes beyond purely perceptive questionnaires. Thus, unconscious reactions to acoustic stimuli can also be captured. The numerous physiological signals and possible experimental designs with acoustic stimuli may quickly lead to a challenging implementation of the study and an increased difficulty in reproduction or comparison between studies. An unsuitable experimental design or processing of the physiological data may result in conclusions about psychological loads that are not valid anymore.Approach. The systematic review according to the preferred reporting items for systematic reviews and meta-analysis standard presented here is therefore intended to provide guidance and a basis for further studies in this field. For this purpose, studies were identified in which the participants' short-term physiological responses to acoustic stimuli were investigated in the context of a listening test in a laboratory study.Main Results. A total of 37 studies met these criteria and data items were analysed in terms of the experimental design (studied psychological load, independent variables/acoustic stimuli, participants, playback, scenario/context, duration of test phases, questionnaires for perceptual comparison) and the physiological signals (measures, calculated features, systems, data processing methods, data analysis methods, results). The overviews show that stress is the most studied psychological load in response to acoustic stimuli. An ECG/PPG system and the measurement of skin conductance were most frequently used for the detection of psychological loads. A critical aspect is the numerous different methods of experimental design, which prevent comparability of the results. In the future, more standardized methods are needed to achieve more valid analyses of the effects of acoustic stimuli.

Automatic Generation of Labeled Data for Video-Based Human Pose Analysis via NLP applied to YouTube Subtitles.

With recent advancements in computer vision as well as machine learning (ML), video-based at-home exercise evaluation systems have become a popular topic of current research. However, performance depends heavily on the amount of available training data. Since labeled datasets specific to exercising are rare, we propose a method that makes use of the abundance of fitness videos available online. Specifically, we utilize the advantage that videos often not only show the exercises, but also provide language as an additional source of information. With push-ups as an example, we show that through the analysis of subtitle data using natural language processing (NLP), it is possible to create a labeled (irrelevant, relevant correct, relevant incorrect) dataset containing relevant information for pose analysis. In particular, we show that irrelevant clips (n = 332) have significantly different joint visibility values compared to relevant clips (n = 298). Inspecting cluster centroids also show different poses for the different classes.

Copula-Based Data Augmentation on a Deep Learning Architecture for Cardiac Sensor Fusion.

In the wake of Big Data, traditional Machine Learning techniques are now often integrated in the clinical workflow. Despite more capable, Deep Learning methods are not equally accepted given their unsatiated need for great amounts of training data and transversal use of the same architectures in fundamentally different areas with weakly-substantiated adaptations. To address the former, a cardiorespiratory signal synthesizer was designed by conditional sampling from a multimodally trained stochastic system of Gaussian copulas integrated in a Markov chain. With respect to the latter, a multi-branch convolutional neural network architecture was conceived to learn the best cardiac sensor-fusion strategy at every abstraction layer. The network was tailored to the tasks of cycle detection and classification for different cardiac modality combinations by a synthesizer-based data augmentation training framework and Bayesian hyperparameter optimization. The synthesizer yielded highly realistic signals in the time, frequency and phase domains for both healthy and pathological heart cycles as well as artifacts of different modalities. Benchmarking suggested that the network is able to surpass previous architectures and data augmentation provided a performance boost in realistic data availability scenarios. These included insufficient training data volume, as low as 150 cycles long, artifact contamination and absence of a classification data type in training.

Noncontact Monitoring of Heart Rate and Heart Rate Variability in Geriatric Patients Using Photoplethysmography Imaging.

OBJECTIVE: Geriatric patients, especially those with dementia or in a delirious state, do not accept conventional contact-based monitoring. Therefore, we propose to measure heart rate (HR) and heart rate variability (HRV) of geriatric patients in a noncontact and unobtrusive way using photoplethysmography imaging (PPGI). METHODS: PPGI video sequences were recorded from 10 geriatric patients and 10 healthy elderly people using a monochrome camera operating in the near-infrared spectrum and a colour camera operating in the visible spectrum. PPGI waveforms were extracted from both cameras using superpixel-based regions of interests (ROI). A classifier based on bagged trees was trained to automatically select artefact-free ROIs for HR estimation. HRV was calculated in the time-domain and frequency-domain. RESULTS: an RMSE of 1.03 bpm and a correlation of 0.8 with the reference was achieved using the NIR camera for HR estimation. Using the RGB camera, RMSE and correlation improved to 0.48 bpm and 0.95, respectively. Correlation for HRV in the frequency-domain (LF/HF-ratio) was 0.50 using the NIR camera and 0.70 using the RGB camera. CONCLUSION: We were able to demonstrate that PPGI is very suitable to measure HR and HRV in geriatric patients. We strongly believe that PPGI will become clinically relevant in monitoring of geriatric patients. SIGNIFICANCE: we are the first group to measure both HR and HRV in awake geriatric patients using PPGI. Moreover, we systematically evaluate the effects of the spectrum (near-infrared vs. visible), ROI, and additional motion artefact reduction algorithms on the accuracy of estimated HR and HRV.

A Neonatal Phantom for Vital Signs Simulation.

Neonatal intensive care units provide vital medical support for premature infants. The key aspect in neonatal care is the continuous monitoring of vital signs measured using adhesive skin sensors. Since sensors can cause irritation of the skin and lead to infections, research focuses on contact-free, camera-based methods such as infrared thermography and photoplethysmography imaging. The development of image processing algorithms requires large datasets, but recording the necessary data from studies brings tremendous effort and costs. Therefore, realistic patient phantoms would be feasible to create a comprehensive dataset and validate image-based algorithms. This work describes the realization of a neonatal phantom which can simulate physiological vital parameters such as pulse rate and thermoregulation. It mimics the outer appearance of premature infants using a 3D printed base structure coated with several layers of modified, skin-colored silicone. A distribution of red and infrared LEDs in the scaffold enables the simulation of a PPG signal by mimicking pulsative light intensity changes on the skin. Additionally, the body temperature of the phantom is individually adjustable in several regions using heating elements. In the validation process for PPG simulation, the feasibility of setting different pulse frequencies and the variation of oxygen saturation levels was obtained. Furthermore, heating tests showed region-dependent temperature variations between 0.19 °C and 0.81 °C around the setpoint. In conclusion, the proposed neonatal phantom can be used to simulate a variety of vital parameters of preterm infants and, therefore, enables the implementation of image processing algorithms for the analysis of the medical state.

Photoplethysmography imaging:camera performance evaluation by means of an optoelectronic skin perfusion phantom.

OBJECTIVE: Photoplethysmography imaging (PPGI) is a promising contactless camera-based method of non-invasive cardiovascular diagnostics. To achieve the best results, it is important to choose the most suitable camera for a specific application. The settings of the camera influence the quality of the detected signal. APPROACH: The standard (European Machine Vision Association 2016 EMVA Standard 1288-Standard for Characterization of Image Sensors and Cameras pp 1-39 (available at: https://www.emva.org/wp-content/uploads/EMVA1288-3.1a.pdf)) for evaluating the imaging performance of machine vision cameras (MVC) helps at the initial decision of the sensor, but the camera should always be tested in terms of usability for a specific application. So far, PPGI lacks a standardized measurement scenario for evaluating the performance of individual cameras. For this, we realized a controllable optoelectronic phantom with artificial silicone skin allowing reproducible tests of cameras to verify their suitability for PPGI. The entire system is housed in a light-tight box. We tested an MVC, a digital single-lens reflex camera (DSLR) camera and a webcam. Each camera varies in used technology and price. MAIN RESULTS: We simulated real PPGI measurement conditions simulating the ratio of pulse (AC) and non-pulse (DC) component of the photoplethysmographic signal and achieved AC/DC ratios of 0.5 % on average. An additional OLED panel ensures proper DC providing reproducible measurement conditions. We evaluated the signal morphological features, amplitude spectrum, signal-to-noise ratio (SNR) and spatially dependent changes of simulated subcutaneous perfusion. Here, the MVC proved to be the most suitable device. A DSLR is also suitable for PPGI, but a larger smoothing kernel is required to obtain a perfusion map. The webcam, as the weakest contender, proved to be very susceptible to any inhomogeneous illumination of the examined artificial skin surface. However, it is still able to detect cardiac rhythm. SIGNIFICANCE: The result of our work is an optoelectronic phantom for reproducible testing of PPGI camera performance in terms of signal quality and measurement equipment costs.

Fast body part segmentation and tracking of neonatal video data using deep learning.

Photoplethysmography imaging (PPGI) for non-contact monitoring of preterm infants in the neonatal intensive care unit (NICU) is a promising technology, as it could reduce medical adhesive-related skin injuries and associated complications. For practical implementations of PPGI, a region of interest has to be detected automatically in real time. As the neonates' body proportions differ significantly from adults, existing approaches may not be used in a straightforward way, and color-based skin detection requires RGB data, thus prohibiting the use of less-intrusive near-infrared (NIR) acquisition. In this paper, we present a deep learning-based method for segmentation of neonatal video data. We augmented an existing encoder-decoder semantic segmentation method with a modified version of the ResNet-50 encoder. This reduced the computational time by a factor of 7.5, so that 30 frames per second can be processed at 960 × 576 pixels. The method was developed and optimized on publicly available databases with segmentation data from adults. For evaluation, a comprehensive dataset consisting of RGB and NIR video recordings from 29 neonates with various skin tones recorded in two NICUs in Germany and India was used. From all recordings, 643 frames were manually segmented. After pre-training the model on the public adult data, parts of the neonatal data were used for additional learning and left-out neonates are used for cross-validated evaluation. On the RGB data, the head is segmented well (82% intersection over union, 88% accuracy), and performance is comparable with those achieved on large, public, non-neonatal datasets. On the other hand, performance on the NIR data was inferior. By employing data augmentation to generate additional virtual NIR data for training, results could be improved and the head could be segmented with 62% intersection over union and 65% accuracy. The method is in theory capable of performing segmentation in real time and thus it may provide a useful tool for future PPGI applications. Graphical Abstract This work presents the development of a customized, real-time capable Deep Learning architecture for segmenting of neonatal videos recorded in the intensive care unit. In addition to hand-annotated data, transfer learning is exploited to improve performance.

Using a Motion Capture System as Reference for Motion Tracking in Photoplethysmography Imaging.

Photoplethysmography Imaging (PPGI) is a camera-based and non-contact technology for measurement of physiological signals. It has been shown that important physiological parameters such as heart rate, heart rate variability and respiratory rate can be derived from PPGI. However, as is the case with most non-contact measurement techniques, motion artefacts present a major challenge. Various algorithms for application to both the 2D PPGI video frames as well as the resulting 1D PPGI waveforms have been developed in order to enhance robustness against motion. In this paper, we focus on the aspect of feature point tracking in the 2D PPGI video sequences. We present an experimental setup, where we used a motion capture system in order to obtain a reference for motion during the recording of PPGI video sequences. In a laboratory experiment, PPGI video sequences were recorded from ten healthy volunteers, who were asked to perform various movements during the recording. The KLT tracking algorithm was applied to the recorded sequences and results compared with the reference values from the motion capture system. The results indicate, that tracking of measurement regions in PPGI video sequences is only one element towards motion robust PPGI. In most scenarios, tracking is not sufficiently precise, requiring further processing of the PPGI waveforms in order to reduce motion artefacts in PPGI signals. These indications were confirmed by further analysis when we looked into the effects of tracking on PPGI heart rate extraction.

A Broader Look: Camera-Based Vital Sign Estimation across the Spectrum.

OBJECTIVES: Camera-based vital sign estimation allows the contactless assessment of important physiological parameters. Seminal contributions were made in the 1930s, 1980s, and 2000s, and the speed of development seems ever increasing. In this suivey, we aim to overview the most recent works in this area, describe their common features as well as shortcomings, and highlight interesting "outliers". METHODS: We performed a comprehensive literature research and quantitative analysis of papers published between 2016 and 2018. Quantitative information about the number of subjects, studies with healthy volunteers vs. pathological conditions, public datasets, laboratory vs. real-world works, types of camera, usage of machine learning, and spectral properties of data was extracted. Moreover, a qualitative analysis of illumination used and recent advantages in terms of algorithmic developments was also performed. RESULTS: Since 2016, 116 papers were published on camera-based vital sign estimation and 59% of papers presented results on 20 or fewer subjects. While the average number of participants increased from 15.7 in 2016 to 22.9 in 2018, the vast majority of papers (n=100) were on healthy subjects. Four public datasets were used in 10 publications. We found 27 papers whose application scenario could be considered a real-world use case, such as monitoring during exercise or driving. These include 16 papers that dealt with non-healthy subjects. The majority of papers (n=61) presented results based on visual, red-green-blue (RGB) information, followed by RGB combined with other parts of the electromagnetic spectrum (n=18), and thermography only (n=12), while other works (n=25) used other mono- or polychromatic non-RGB data. Surprisingly, a minority of publications (n=39) made use of consumer-grade equipment. Lighting conditions were primarily uncontrolled or ambient. While some works focused on specialized aspects such as the removal of vital sign information from video streams to protect privacy or the influence of video compression, most algorithmic developments were related to three areas: region of interest selection, tracking, or extraction of a one-dimensional signal. Seven papers used deep learning techniques, 17 papers used other machine learning approaches, and 92 made no explicit use of machine learning. CONCLUSION: Although some general trends and frequent shortcomings are obvious, the spectrum of publications related to camera-based vital sign estimation is broad. While many creative solutions and unique approaches exist, the lack of standardization hinders comparability of these techniques and of their performance. We believe that sharing algorithms and/ or datasets will alleviate this and would allow the application of newer techniques such as deep learning.

Modeling photoplethysmographic signals in camera-based perfusion measurements: optoelectronic skin phantom.

The remote acquisition of photoplethysmographic (PPG) signals via a video camera, also known as photoplethysmography imaging (PPGI), is not yet standardized. In general, PPGI is investigated with test persons in a laboratory setting. While these in-vivo tests have the advantage of generating real-life data, they suffer from the lack of repeatability and are comparatively effort-intensive because human subjects are required. Consequently, studying changes in signal morphology, for example, due to aging or pathological effects, is practically impossible. As a tool to study these effects, a hardware PPG simulator has been developed: this is a phantom which simulates and generates both 1D and locally resolved 2D optical PPG signals. Here, we demonstrate that it is possible to generate PPG-like signals with various signal morphologies by means of a purely optoelectronic setup, namely an LED array, and to analyze them by means of PPGI. Signals extracted via a camera show good agreement with simulated generated signals. In fact, the first phantom design is suitable to demonstrate this qualitatively.

Waveform Analysis for Camera-based Photoplethysmography Imaging.

With the advent of sensitive and affordable cameras, classical contact-based photoplethysmography (PPG) could be enhanced to the spatial domain. Cost-efficient cameras are available in everyday items such as smartphones or computer webcams. The PPG signal, blood volume changes in the vascularity, can be measured remotely by using the camera as a 2-D-PPG detector. However, the evaluation of the extracted signals has mostly been limited to the pulse rate and sometimes the systolic amplitude. In this work, we motivate to generate images and video sequences based on features from the PPG waveform commonly not extracted via cameras. This is achieved by calculating the features for timeseries extracted from an evenly spaced grid of virtual PPG sensors. We briefly discuss the adaption of conventional PPG algorithms to camera-based PPG imaging (PPGI). The extracted parameters are associated with vessel properties and thus, mapping these to images could lead to enhanced vascular diagnostics. In this work, we test the feasibility of the mapping approach: we present the preliminary results gathered from the analysis of two videos of lab experiments with healthy subjects.

A Multi-Modal Sensor for a Bed-Integrated Unobtrusive Vital Signs Sensing Array.

In this paper, we present a novel unobtrusive multi-modal sensor for monitoring of physiological parameters featuring capacitive electrocardiogram (cECG), reflective photoplethysmogram (rPPG), and magnetic induction monitoring (MI) in a single sensor. The sensor system comprises sensor nodes designed and optimized for integration into a grid-like array of multiple sensors in a bed and a central controller box for data collection and processing. Hence, it is highly versatile in application and suitable for unobtrusive monitoring of vital signs, both in a professional setting and a home-care environment. The presented hardware design takes both inter-modal interference between cECG and MI into account as well as intra-modal interference due to cross talk between two MI sensors in close vicinity. In a lab study, we evaluated a prototype of our new multi-modal sensor with two sensor nodes on four healthy subjects. The subjects were lying on the sensors and exercising with a hand grip in order to increase heart rate and thus evaluate our sensor both during changing physiological parameters as well as a wider range of those. Heart beat intervals and heart rate variability were derived from both cECG and rPPG. Breathing intervals were derived from the MI sensor. For heart beat intervals, we achieved an RMSE of 2.3 ms and a correlation of 0.99 using cECG. Similarly, using rPPG, an RMSE of 18.9 ms with a correlation of 0.99 was achieved. With regard to breathing intervals derived from MI, we achieved an RMSE of 1.12 s and a correlation of 0.90.

Local Interval Estimation Improves Accuracy and Robustness of Heart Rate Variability Derivation from Photoplethysmography.

Heart rate variability (HRV) can contain useful information about a subject, but its derivation traditionally relies on conductive electrocardiography (ECG) with adhesive electrodes. While photoplethysmography (PPG) can be acquired in much less intrusive ways, its signal differs fundamentally from ECG. First, it represents mechanical cardiac activity instead of electrical. Second, fiducial points of its waveform are much smoother compared to the QRS complex of the ECG. Still, studies have shown that meaningful HRV parameters can be extracted using PPG which small differences compared to ECG. In this work, we evaluate an algorithm termed "continuous local interval estimator (CLIE)" that analyzes the signal's entire waveform instead of individual fiducial points with respect to its potential in deriving beat-to-beat intervals and the time-domain HRV parameters SDNN, RMSSD, and pNN50 from the PPG. For evaluation, a polysomnography dataset consisting of more than 900,000 recorded heart beats from 33 subjects was used. The performance of CLIE was compared to three peak-detection strategies (peak-to-peak, peak-to-peak of first derivative, troth-to-troth) often found in the literature. For interval estimation and the proposed HRV parameters, CLIE outperformed the reference methods in terms of accuracy. Moreover, when the signal was contaminated with simulated noise, the performance of CLIE was affected only minimally compared to the references. While an adaptive prior could increase the performance of CLIE for very noisy signals, its application was found to deteriorate results when no noise was added. Thus, CLIE was found to be an accurate and robust tool when deriving HRV parameters from PPG signals, which can be augmented by an adaptive prior for potentially noisy signals, such as PPG imaging or wearable PPG.

Modelling and Synthesizing Motion Artifacts in Unobtrusive Multimodal Sensing using Copulas.

The use of non-contact sensing modalities to estimate apatient's vital signs is a promising approach to improve remote monitoring. One of the main challenges in non-contact sensing are motion artifacts, which can cause severe problems and must not be disregarded when designing non-contact systems. Combining multiple sensors and using intelligent sensor-fusion algorithms can reduce the influence of motion artifacts and improve the robustness of the vital sign estimation. Training and validating algorithms are important parts of the development process, but acquiring real data is usually a time-consuming task. Therefore a method to generate a large number of multi-sensor motion artifacts is needed. In this paper we investigate motion artifacts and their inter-dependence in a multi-sensor system. From these analyses, a multivariate mathematical artifact model is derived. Further-more, we propose a general synthesizing algorithm for artificial motion artifacts that allows creating an arbitrary number of multi-sensor motion artifacts. Finally, we compare the artificially created artifacts with real artifacts and evaluate our algorithm. Both qualitative indicators, e.g. signal morphology, and quantitative analyses, e.g. statistical distance measures, show a good accuracy of our model.

Non-Contact Remote Measurement of Heart Rate Variability using Near-Infrared Photoplethysmography Imaging.

Heart rate variability (HRV) is an important clinical parameter associated with the autonomous nervous system (ANS), age, as well as many diseases such as myocardial infarction, diabetes or renal failure. Gold standard for measurement of HRV is a high-resolution electrocardiogram (ECG). With the current trend towards non-contact and unobtrusive monitoring of vital signs, HRV has also become an interesting and important parameter for non-contact monitoring. In this paper, we present an approach towards non-contact and unobtrusive monitoring of heart rate variability using the camera-based technology of photoplethysmography imaging (PPGI). We investigated the suitability of invisible near-infrared illumination for PPGI, which would enable measurement of HRV in darkness. We compared results obtained using infrared illumination with those obtained using visible light as PPGI illumination and calculated both time-domain as well as frequency-domain HRV parameters. The results achieved with infrared illumination were on par with those using conventional illumination in the visible spectrum. We concluded that infrared illumination enables unobtrusive and non-contact remote HRV measurement in both darkness as well as regular daylight conditions using PPGI.

MuSeSe - A multisensor armchair for unobtrusive vital sign estimation and motion artifact analysis.

Unobtrusive vital sign estimation with sensors integrated into objects of everyday living can substantially advance the field of remote monitoring. At the same time, motion artifacts cause severe problems and have to be dealt with. Here, the fusion of multimodal sensor data is a promising approach. In this paper, we present an armchair equipped with capacitively coupled electrocardiogram, two types of ballistocardiographic sensors, photoplethysmographic and two high-frequency impedance sensors. In addition, a video-based sensor for motion analysis is integrated. Using a defined motion protocol, the feasibility of the system is demonstrated in a self-experimentation. Moreover, the influence of different movements on different modalities is analyzed. Finally, robust beat-to-beat interval estimation demonstrates the benefits of multimodal sensor fusion for vital sign estimation in the presence of motion artifacts.

Reducing false alarms in the ICU by quantifying self-similarity of multimodal biosignals.

False arrhythmia alarms pose a major threat to the quality of care in today's ICU. Thus, the PhysioNet/Computing in Cardiology Challenge 2015 aimed at reducing false alarms by exploiting multimodal cardiac signals recorded by a patient monitor. False alarms for asystole, extreme bradycardia, extreme tachycardia, ventricular flutter/fibrillation as well as ventricular tachycardia were to be reduced using two electrocardiogram channels, up to two cardiac signals of mechanical origin as well as a respiratory signal. In this paper, an approach combining multimodal rhythmicity estimation and machine learning is presented. Using standard short-time autocorrelation and robust beat-to-beat interval estimation, the signal's self-similarity is analyzed. In particular, beat intervals as well as quality measures are derived which are further quantified using basic mathematical operations (min, mean, max, etc). Moreover, methods from the realm of image processing, 2D Fourier transformation combined with principal component analysis, are employed for dimensionality reduction. Several machine learning approaches are evaluated including linear discriminant analysis and random forest. Using an alarm-independent reduction strategy, an overall false alarm reduction with a score of 65.52 in terms of the real-time scoring system of the challenge is achieved on a hidden dataset. Employing an alarm-specific strategy, an overall real-time score of 78.20 at a true positive rate of 95% and a true negative rate of 78% is achieved. While the results for some categories still need improvement, false alarms for extreme tachycardia are suppressed with 100% sensitivity and specificity.

Detection of heart beats in multimodal data: a robust beat-to-beat interval estimation approach.

The heart rate and its variability play a vital role in the continuous monitoring of patients, especially in the critical care unit. They are commonly derived automatically from the electrocardiogram as the interval between consecutive heart beat. While their identification by QRS-complexes is straightforward under ideal conditions, the exact localization can be a challenging task if the signal is severely contaminated with noise and artifacts. At the same time, other signals directly related to cardiac activity are often available. In this multi-sensor scenario, methods of multimodal sensor-fusion allow the exploitation of redundancies to increase the accuracy and robustness of beat detection.In this paper, an algorithm for the robust detection of heart beats in multimodal data is presented. Classic peak-detection is augmented by robust multi-channel, multimodal interval estimation to eliminate false detections and insert missing beats. This approach yielded a score of 90.70 and was thus ranked third place in the PhysioNet/Computing in Cardiology Challenge 2014: Robust Detection of Heart Beats in Muthmodal Data follow-up analysis.In the future, the robust beat-to-beat interval estimator may directly be used for the automated processing of multimodal patient data for applications such as diagnosis support and intelligent alarming.