Deep learning in spatially resolved transcriptfomics: a comprehensive technical view

Spatially resolved transcriptomics (SRT) is a pioneering method for simultaneously studying morphological contexts and gene expression at single-cell precision. Data emerging from SRT are multifaceted, presenting researchers with intricate gene expression matrices, precise spatial details and comprehensive histology visuals. Such rich and intricate datasets, unfortunately, render many conventional methods like traditional machine learning and statistical models ineffective. The unique challenges posed by the specialized nature of SRT data have led the scientific community to explore more sophisticated analytical avenues. Recent trends indicate an increasing reliance on deep learning algorithms, especially in areas such as spatial clustering, identification of spatially variable genes and data alignment tasks. In this manuscript, we provide a rigorous critique of these advanced deep learning methodologies, probing into their merits, limitations and avenues for further refinement. Our in-depth analysis underscores that while the recent innovations in deep learning tailored for SRT have been promising, there remains a substantial potential for enhancement. A crucial area that demands attention is the development of models that can incorporate intricate biological nuances, such as phylogeny-aware processing or in-depth analysis of minuscule histology image segments. Furthermore, addressing challenges like the elimination of batch effects, perfecting data normalization techniques and countering the overdispersion and zero inflation patterns seen in gene expression is pivotal. To support the broader scientific community in their SRT endeavors, we have meticulously assembled a comprehensive directory of readily accessible SRT databases, hoping to serve as a foundation for future research initiatives.

Non-invasive parameters of autonomic function using beat-to-beat cardiovascular variations and arterial stiffness in hypertensive individuals: a systematic review.

PURPOSE: Non-invasive, beat-to-beat variations in physiological indices provide an opportunity for more accessible assessment of autonomic dysfunction. The potential association between the changes in these parameters and arterial stiffness in hypertension remains poorly understood. This systematic review aims to investigate the association between non-invasive indicators of autonomic function based on beat-to-beat cardiovascular signals with arterial stiffness in individuals with hypertension. METHODS: Four electronic databases were searched from inception to June 2022. Studies that investigated non-invasive parameters of arterial stiffness and autonomic function using beat-to-beat cardiovascular signals over a period of > 5min were included. Study quality was assessed using the STROBE criteria. Two authors screened the titles, abstracts, and full texts independently. RESULTS: Nineteen studies met the inclusion criteria. A comprehensive overview of experimental design for assessing autonomic function in terms of baroreflex sensitivity and beat-to-beat cardiovascular variabilities, as well as arterial stiffness, was presented. Alterations in non-invasive indicators of autonomic function, which included baroreflex sensitivity, beat-to-beat cardiovascular variabilities and hemodynamic changes in response to autonomic challenges, as well as arterial stiffness, were identified in individuals with hypertension. A mixed result was found in terms of the association between non-invasive quantitative autonomic indices and arterial stiffness in hypertensive individuals. Nine out of 12 studies which quantified baroreflex sensitivity revealed a significant association with arterial stiffness parameters. Three studies estimated beat-to-beat heart rate variability and only one study reported a significant relationship with arterial stiffness indices. Three out of five studies which studied beat-to-beat blood pressure variability showed a significant association with arterial structural changes. One study revealed that hemodynamic changes in response to autonomic challenges were significantly correlated with arterial stiffness parameters. CONCLUSIONS: The current review demonstrated alteration in autonomic function, which encompasses both the sympathetic and parasympathetic modulation of sinus node function and vasomotor tone (derived from beat-to-beat cardiovascular signals) in hypertension, and a significant association between some of these parameters with arterial stiffness. By employing non-invasive measurements to monitor changes in autonomic function and arterial remodeling in individuals with hypertension, we would be able to enhance our ability to identify individuals at high risk of cardiovascular disease. Understanding the intricate relationships among these cardiovascular variability measures and arterial stiffness could contribute toward better individualized treatment for hypertension in the future. SYSTEMATIC REVIEW REGISTRATION: PROSPERO ID: CRD42022336703. Date of registration: 12/06/2022.

A Smart, Textile-Driven, Soft Exosuit for Spinal Assistance.

Work-related musculoskeletal disorders (WMSDs) are often caused by repetitive lifting, making them a significant concern in occupational health. Although wearable assist devices have become the norm for mitigating the risk of back pain, most spinal assist devices still possess a partially rigid structure that impacts the user's comfort and flexibility. This paper addresses this issue by presenting a smart textile-actuated spine assistance robotic exosuit (SARE), which can conform to the back seamlessly without impeding the user's movement and is incredibly lightweight. To detect strain on the spine and to control the smart textile automatically, a soft knitting sensor that utilizes fluid pressure as a sensing element is used. Based on the soft knitting hydraulic sensor, the robotic exosuit can also feature the ability of monitoring and rectifying human posture. The SARE is validated experimentally with human subjects (N = 4). Through wearing the SARE in stoop lifting, the peak electromyography (EMG) signals of the lumbar erector spinae are reduced by 22.8% ± 12 for lifting 5 kg weights and 27.1% ± 14 in empty-handed conditions. Moreover, the integrated EMG decreased by 34.7% ± 11.8 for lifting 5 kg weights and 36% ± 13.3 in empty-handed conditions. In summary, the artificial muscle wearable device represents an anatomical solution to reduce the risk of muscle strain, metabolic energy cost and back pain associated with repetitive lifting tasks.

KARAJ: An Efficient Adaptive Multi-Processor Tool to Streamline Genomic and Transcriptomic Sequence Data Acquisition.

Here we developed KARAJ, a fast and flexible Linux command-line tool to automate the end-to-end process of querying and downloading a wide range of genomic and transcriptomic sequence data types. The input to KARAJ is a list of PMCIDs or publication URLs or various types of accession numbers to automate four tasks as follows; firstly, it provides a summary list of accessible datasets generated by or used in these scientific articles, enabling users to select appropriate datasets; secondly, KARAJ calculates the size of files that users want to download and confirms the availability of adequate space on the local disk; thirdly, it generates a metadata table containing sample information and the experimental design of the corresponding study; and lastly, it enables users to download supplementary data tables attached to publications. Further, KARAJ provides a parallel downloading framework powered by Aspera connect which reduces the downloading time significantly.

Quantitative neurogenetics: applications in understanding disease.

Neurodevelopmental and neurodegenerative disorders (NNDs) are a group of conditions with a broad range of core and co-morbidities, associated with dysfunction of the central nervous system. Improvements in high throughput sequencing have led to the detection of putative risk genetic loci for NNDs, however, quantitative neurogenetic approaches need to be further developed in order to establish causality and underlying molecular genetic mechanisms of pathogenesis. Here, we discuss an approach for prioritizing the contribution of genetic risk loci to complex-NND pathogenesis by estimating the possible impacts of these loci on gene regulation. Furthermore, we highlight the use of a tissue-specificity gene expression index and the application of artificial intelligence (AI) to improve the interpretation of the role of genetic risk elements in NND pathogenesis. Given that NND symptoms are associated with brain dysfunction, risk loci with direct, causative actions would comprise genes with essential functions in neural cells that are highly expressed in the brain. Indeed, NND risk genes implicated in brain dysfunction are disproportionately enriched in the brain compared with other tissues, which we refer to as brain-specific expressed genes. In addition, the tissue-specificity gene expression index can be used as a handle to identify non-brain contexts that are involved in NND pathogenesis. Lastly, we discuss how using an AI approach provides the opportunity to integrate the biological impacts of risk loci to identify those putative combinations of causative relationships through which genetic factors contribute to NND pathogenesis.

A Wearable Soft Fabric Sleeve for Upper Limb Augmentation.

Soft actuators (SAs) have been used in many compliant robotic structure and wearable devices, due to their safe interaction with the wearers. Despite advances, the capability of current SAs is limited by scalability, high hysteresis, and slow responses. In this paper, a new class of soft, scalable, and high-aspect ratio fiber-reinforced hydraulic SAs is introduced. The new SA uses a simple fabrication process of insertion where a hollow elastic rubber tube is directly inserted into a constrained hollow coil, eliminating the need for the manual wrapping of an inextensible fiber around a long elastic structure. To provide high adaptation to the user skin for wearable applications, the new SAs are integrated into flexible fabrics to form a wearable fabric sleeve. To monitor the SA elongation, a soft liquid metal-based fabric piezoresistive sensor is also developed. To capture the nonlinear hysteresis of the SA, a novel asymmetric hysteresis model which only requires five model parameters in its structure is developed and experimentally validated. The new SAs-driven wearable robotic sleeve is scalable, highly flexible, and lightweight. It can also produce a large amount of force of around 23 N per muscle at around 30% elongation, to provide useful assistance to the human upper limbs. Experimental results show that the soft fabric sleeve can augment a user's performance when working against a load, evidenced by a significant reduction on the muscular effort, as monitored by electromyogram (EMG) signals. The performance of the developed SAs, soft fabric sleeve, soft liquid metal fabric sensor, and nonlinear hysteresis model reveal that they can effectively modulate the level of assistance for the wearer. The new technologies obtained from this work can be potentially implemented in emerging assistive applications, such as rehabilitation, defense, and industry.

Estimating Lower Limb Kinematics Using a Lie Group Constrained Extended Kalman Filter with a Reduced Wearable IMU Count and Distance Measurements.

Tracking the kinematics of human movement usually requires the use of equipment that constrains the user within a room (e.g., optical motion capture systems), or requires the use of a conspicuous body-worn measurement system (e.g., inertial measurement units (IMUs) attached to each body segment). This paper presents a novel Lie group constrained extended Kalman filter to estimate lower limb kinematics using IMU and inter-IMU distance measurements in a reduced sensor count configuration. The algorithm iterates through the prediction (kinematic equations), measurement (pelvis height assumption/inter-IMU distance measurements, zero velocity update for feet/ankles, flat-floor assumption for feet/ankles, and covariance limiter), and constraint update (formulation of hinged knee joints and ball-and-socket hip joints). The knee and hip joint angle root-mean-square errors in the sagittal plane for straight walking were 7.6±2.6∘ and 6.6±2.7∘, respectively, while the correlation coefficients were 0.95±0.03 and 0.87±0.16, respectively. Furthermore, experiments using simulated inter-IMU distance measurements show that performance improved substantially for dynamic movements, even at large noise levels (σ=0.2 m). However, further validation is recommended with actual distance measurement sensors, such as ultra-wideband ranging sensors.

Deep Learning for Activity Recognition in Older People Using a Pocket-Worn Smartphone.

Activity recognition can provide useful information about an older individual's activity level and encourage older people to become more active to live longer in good health. This study aimed to develop an activity recognition algorithm for smartphone accelerometry data of older people. Deep learning algorithms, including convolutional neural network (CNN) and long short-term memory (LSTM), were evaluated in this study. Smartphone accelerometry data of free-living activities, performed by 53 older people (83.8 ± 3.8 years; 38 male) under standardized circumstances, were classified into lying, sitting, standing, transition, walking, walking upstairs, and walking downstairs. A 1D CNN, a multichannel CNN, a CNN-LSTM, and a multichannel CNN-LSTM model were tested. The models were compared on accuracy and computational efficiency. Results show that the multichannel CNN-LSTM model achieved the best classification results, with an 81.1% accuracy and an acceptable model and time complexity. Specifically, the accuracy was 67.0% for lying, 70.7% for sitting, 88.4% for standing, 78.2% for transitions, 88.7% for walking, 65.7% for walking downstairs, and 68.7% for walking upstairs. The findings indicated that the multichannel CNN-LSTM model was feasible for smartphone-based activity recognition in older people.

Learning the Orientation of a Loosely-Fixed Wearable IMU Relative to the Body Improves the Recognition Rate of Human Postures and Activities.

Features were developed which accounted for the changing orientation of the inertial measurement unit (IMU) relative to the body, and demonstrably improved the performance of models for human activity recognition (HAR). The method is proficient at separating periods of standing and sedentary activity (i.e., sitting and/or lying) using only one IMU, even if it is arbitrarily oriented or subsequently re-oriented relative to the body; since the body is upright during walking, learning the IMU orientation during walking provides a reference orientation against which sitting and/or lying can be inferred. Thus, the two activities can be identified (irrespective of the cohort) by analyzing the magnitude of the angle of shortest rotation which would be required to bring the upright direction into coincidence with the average orientation from the most recent 2.5 s of IMU data. Models for HAR were trained using data obtained from a cohort of 37 older adults (83.9 ± 3.4 years) or 20 younger adults (21.9 ± 1.7 years). Test data were generated from the training data by virtually re-orienting the IMU so that it is representative of carrying the phone in five different orientations (relative to the thigh). The overall performance of the model for HAR was consistent whether the model was trained with the data from the younger cohort, and tested with the data from the older cohort after it had been virtually re-oriented (Cohen's Kappa 95% confidence interval [0.782, 0.793]; total class sensitivity 95% confidence interval [84.9%, 85.6%]), or the reciprocal scenario in which the model was trained with the data from the older cohort, and tested with the data from the younger cohort after it had been virtually re-oriented (Cohen's Kappa 95% confidence interval [0.765, 0.784]; total class sensitivity 95% confidence interval [82.3%, 83.7%]).

Long-term anesthetic protocol in rats: feasibility in electrophysiology studies in visual prosthesis.

Electrical stimulation of excitable cells provides therapeutic benefits for a variety of medical conditions, including restoration of partial vision to those blinded via some types of retinal degeneration. To improve visual percepts elicited by the current technology, researchers are conducting acute electrophysiology experiments, mainly in cats. However, the rat can provide a model of a range of retinal diseases and possesses a sufficiently large eye to be used in this field. This article presents a long-term anesthetic protocol to enable electrophysiology experiments to further the development of visual prostheses. Six Long-Evans rats (aged between 14 and 16 weeks) were included in this study. Surgical anesthesia was maintained for more than 15 h by combining constant intravenous infusion of ketamine (24.0-34.5 mg/kg/h), xylazine (0.9-1.2 mg/kg/h), and inhaled isoflurane in oxygen (<0.5%). Overall heart rate, respiratory rate, and body temperature remained between 187-233 beats/min, 45-58 breaths/min, and 36-38 °C, respectively. Neural responses to 200-ms light pulses were recorded from the superior colliculus using a 32-channel neural probe at the beginning and before termination of the experiment. Robust responses were recorded from distinct functional types of retinal pathways. In addition, a platinum electrode was implanted in the retrobulbar space. The retina was electrically stimulated, and the activation threshold was determined to be 5.24 ± 0.24 µC/cm2 . This protocol may be used not only in the field of visual prosthesis research, but in other research areas requiring longer term acute experiments.

Differential effect of brief electrical stimulation on voltage-gated potassium channels.

Electrical stimulation of neuronal tissue is a promising strategy to treat a variety of neurological disorders. The mechanism of neuronal activation by external electrical stimulation is governed by voltage-gated ion channels. This stimulus, typically brief in nature, leads to membrane potential depolarization, which increases ion flow across the membrane by increasing the open probability of these voltage-gated channels. In spiking neurons, it is activation of voltage-gated sodium channels (NaV channels) that leads to action potential generation. However, several other types of voltage-gated channels are expressed that also respond to electrical stimulation. In this study, we examine the response of voltage-gated potassium channels (KV channels) to brief electrical stimulation by whole cell patch-clamp electrophysiology and computational modeling. We show that nonspiking amacrine neurons of the retina exhibit a large variety of responses to stimulation, driven by different KV-channel subtypes. Computational modeling reveals substantial differences in the response of specific KV-channel subtypes that is dependent on channel kinetics. This suggests that the expression levels of different KV-channel subtypes in retinal neurons are a crucial predictor of the response that can be obtained. These data expand our knowledge of the mechanisms of neuronal activation and suggest that KV-channel expression is an important determinant of the sensitivity of neurons to electrical stimulation.NEW & NOTEWORTHY This paper describes the response of various voltage-gated potassium channels (KV channels) to brief electrical stimulation, such as is applied during prosthetic electrical stimulation. We show that the pattern of response greatly varies between KV channel subtypes depending on activation and inactivation kinetics of each channel. Our data suggest that problems encountered when artificially stimulating neurons such as cessation in firing at high frequencies, or "fading," may be attributed to KV-channel activation.

Local Heterogeneous Electrical Restitution Properties of Rabbit Atria.

INTRODUCTION: This study aims to characterize the regional variability in rate-adaptation in the atria. METHODS AND RESULTS: Action potential (AP) responses to pulses with uniform as well as pseudo-random non-uniform pacing intervals were recorded from rabbit sino-atrial node, right and left atrial pectinate as well as pulmonary vein antrum tissue preparations using conventional intracellular glass microelectrodes. Steady-state restitution curves were reconstructed for various AP waveform metrics. We observed significant variability between the four regions under basal pacing representing the rabbit resting heart rate as well as regional variability in rate-adaptation to increased pacing frequencies. Right-left atrial restitution differences were further confirmed using the non-uniform pacing protocol, with significant differences in AP amplitude, duration (APD) as well as maximum phase 0 depolarization rate restitution curves in response to an identical sequence of non-uniform pacing intervals. In addition, we report regional differences in alternans of AP waveform metrics, over a wide range of pacing frequencies and not simply prior to 1:1 entrainment being lost. We also observed an increase in APD90 along the conduction pathway from the left atrium to pulmonary vein junction. CONCLUSIONS: Our results identified significant regional differences in electrical restitution in the rabbit atria and suggest their dependency on both baseline AP morphology and local intrinsic differences in rate-adaptation. We propose that the atrial heterogeneity in rate-adaptation could contribute to arrhythmogenesis and the greater susceptibility of pulmonary vein myocardial sleeves to ectopic foci and reentrant activity.

Analyzing health insurance claims on different timescales to predict days in hospital.

Health insurers maintain large databases containing information on medical services utilized by claimants, often spanning several healthcare services and providers. Proper use of these databases could facilitate better clinical and administrative decisions. In these data sets, there exists many unequally spaced events, such as hospital visits. However, data mining of temporal data and point processes is still a developing research area and extracting useful information from such data series is a challenging task. In this paper, we developed a time series data mining approach to predict the number of days in hospital in the coming year for individuals from a general insured population based on their insurance claim data. In the proposed method, the data were windowed at four different timescales (bi-monthly, quarterly, half-yearly and yearly) to construct regularly spaced time series features extracted from such events, resulting in four associated prediction models. A comparison of these models indicates models using a half-yearly windowing scheme delivers the best performance on all three populations (the whole population, a senior sub-population and a non-senior sub-population). The superiority of the half-yearly model was found to be particularly pronounced in the senior sub-population. A bagged decision tree approach was able to predict 'no hospitalization' versus 'at least one day in hospital' with a Matthews correlation coefficient (MCC) of 0.426. This was significantly better than the corresponding yearly model, which achieved 0.375 for this group of customers. Further reducing the length of the analysis windows to three or two months did not produce further improvements.

Tracking the Evolution of Smartphone Sensing for Monitoring Human Movement.

Advances in mobile technology have led to the emergence of the "smartphone", a new class of device with more advanced connectivity features that have quickly made it a constant presence in our lives. Smartphones are equipped with comparatively advanced computing capabilities, a global positioning system (GPS) receivers, and sensing capabilities (i.e., an inertial measurement unit (IMU) and more recently magnetometer and barometer) which can be found in wearable ambulatory monitors (WAMs). As a result, algorithms initially developed for WAMs that "count" steps (i.e., pedometers); gauge physical activity levels; indirectly estimate energy expenditure and monitor human movement can be utilised on the smartphone. These algorithms may enable clinicians to "close the loop" by prescribing timely interventions to improve or maintain wellbeing in populations who are at risk of falling or suffer from a chronic disease whose progression is linked to a reduction in movement and mobility. The ubiquitous nature of smartphone technology makes it the ideal platform from which human movement can be remotely monitored without the expense of purchasing, and inconvenience of using, a dedicated WAM. In this paper, an overview of the sensors that can be found in the smartphone are presented, followed by a summary of the developments in this field with an emphasis on the evolution of algorithms used to classify human movement. The limitations identified in the literature will be discussed, as well as suggestions about future research directions.

On Time Domain Analysis of Photoplethysmogram Signals for Monitoring Heat Stress.

There are a limited number of studies on heat stress dynamics during exercise using the photoplethysmogram (PPG) and its second derivative (APG). However, we investigate the most suitable index from short PPG signal recordings for heat stress assessment. The APG waveform consists of a, b, c and d waves in systole and an e wave in diastole. Our preliminary results indicate that the use of the energy of aa area, derived from PPG signals measured from emergency responders in tropical conditions, is promising in determining the heat stress level using 20-s recordings. After examining 14 time domain features using leave-one-out cross-validation, we found that the aa energy extracted from PPG signals is the most informative feature for classifying heat-stressed subjects, with an overall accuracy of 79%. Moreover, the combination of the aa energy with the traditional Sensors 2015, 15 24717 heart rate variability index of heat stress (i.e., the square root of the mean of the squares of the successive aa intervals) improved the heat stress detection to an overall accuracy of 83%.

Improved Measurement of Blood Pressure by Extraction of Characteristic Features from the Cuff Oscillometric Waveform.

We present a novel approach to improve the estimation of systolic (SBP) and diastolic blood pressure (DBP) from oscillometric waveform data using variable characteristic ratios between SBP and DBP with mean arterial pressure (MAP). This was verified in 25 healthy subjects, aged 28 ± 5 years. The multiple linear regression (MLR) and support vector regression (SVR) models were used to examine the relationship between the SBP and the DBP ratio with ten features extracted from the oscillometric waveform envelope (OWE). An automatic algorithm based on relative changes in the cuff pressure and neighbouring oscillometric pulses was proposed to remove outlier points caused by movement artifacts. Substantial reduction in the mean and standard deviation of the blood pressure estimation errors were obtained upon artifact removal. Using the sequential forward floating selection (SFFS) approach, we were able to achieve a significant reduction in the mean and standard deviation of differences between the estimated SBP values and the reference scoring (MLR: mean ± SD = -0.3 ± 5.8 mmHg; SVR and -0.6 ± 5.4 mmHg) with only two features, i.e., Ratio2 and Area3, as compared to the conventional maximum amplitude algorithm (MAA) method (mean ± SD = -1.6 ± 8.6 mmHg). Comparing the performance of both MLR and SVR models, our results showed that the MLR model was able to achieve comparable performance to that of the SVR model despite its simplicity.

Understanding the retina: a review of computational models of the retina from the single cell to the network level.

The vertebrate retina is a clearly organized signal-processing system. It contains more than 60 different types of neurons, arranged in three distinct neural layers. Each cell type is believed to serve unique role(s) in encoding visual information. While we now have a relatively good understanding of the constituent cell types in the retina and some general ideas of their connectivity, with few exceptions, how the retinal circuitry performs computation remains poorly understood. Computational modeling has been commonly used to study the retina from the single cell to the network level. In this article, we begin by reviewing retinal modeling strategies and existing models. We then discuss in detail the significance and limitations of these models, and finally, we provide suggestions for the future development of retinal neural modeling.

Robust aortic valve non-opening detection for different cardiac conditions.

In recent years, extensive studies have been conducted in the area of pumping state detection for implantable rotary blood pumps. However, limited studies have focused on automatically identifying the aortic valve non-opening (ANO) state despite its importance in the development of control algorithms aiming for myocardial recovery. In the present study, we investigated the performance of 14 ANO indices derived from the pump speed waveform using four different types of classifiers, including linear discriminant analysis, logistic regression, back propagation neural network, and k-nearest neighbors (KNN). Experimental measurements from four greyhounds, which take into consideration the variations in cardiac contractility, systemic vascular resistance, and total blood volume were used. By having only two indices, (i) the root mean square value, and (ii) the standard deviation, we were able to achieve an accuracy of 92.8% with the KNN classifier. Further increase of the number of indices to five for the KNN classifier increases the overall accuracy to 94.6%.

A sliding mode-based starling-like controller for implantable rotary blood pumps.

Clinically adequate implementation of physiological control of a rotary left ventricular assist device requires a sophisticated technique such as the recently proposed method based on the Frank-Starling mechanism. In this mechanism, the stroke volume of the heart increases in response to an increase in the volume of blood filling the left ventricle at the end of diastole. To emulate this process, changes in pump speed need to automatically regulate pump flow to ensure that the combined output of the left ventricle and pump match the output of the right ventricle across changing cardiovascular states. In this approach, we exploit the linear relationship between estimated mean pump flow (Q ̅ est) and pump flow pulsatility (PIQp) in a tracking control algorithm based on sliding mode control. The immediate response of the controller was assessed using a lumped parameter model of the cardiovascular system (CVS) and pump from which could be extracted both Q ̅ est and PIQp. Two different perturbations from the resting state in the presence of left ventricular failure were tested. The first was blood loss requiring a reduction in pump flow to match the reduced output from the right ventricle and to avoid the complication of ventricular suction. The second was exercise, requiring an increase in pump flow. The sliding mode controller induced the required changes in Qp within approximately five heart beats in the blood loss simulation and eight heart beats in the exercise simulation without clinically significant transients or steady-state errors.

Energy-Efficient Sleep Apnea Detection Using a Hyperdimensional Computing Framework Based on Wearable Bracelet Photoplethysmography.

OBJECTIVE: Sleep apnea syndrome (SAS) is a common sleep disorder, which has been shown to be an important contributor to major neurocognitive and cardiovascular sequelae. Considering current diagnostic strategies are limited with bulky medical devices and high examination expenses, a large number of cases go undiagnosed. To enable large-scale screening for SAS, wearable photoplethysmography (PPG) technologies have been used as an early detection tool. However, existing algorithms are energy-intensive and require large amounts of memory resources, which are believed to be the major drawbacks for further promotion of wearable devices for SAS detection. METHODS: In this paper, an energy-efficient method of SAS detection based on hyperdimensional computing (HDC) is proposed. Inspired by the phenomenon of chunking in cognitive psychology as a memory mechanism for improving working memory efficiency, we proposed a one-dimensional block local binary pattern (1D-BlockLBP) encoding scheme combined with HDC to preserve dominant dynamical and temporal characteristics of pulse rate signals from wearable PPG devices. RESULTS: Our method achieved 70.17% accuracy in sleep apnea segment detection, which is comparable with traditional machine learning methods. Additionally, our method achieves up to 67× lower memory footprint, 68× latency reduction, and 93× energy saving on the ARM Cortex-M4 processor. CONCLUSION: The simplicity of hypervector operations in HDC and the novel 1D-BlockLBP encoding effectively preserve pulse rate signal characteristics with high computational efficiency. SIGNIFICANCE: This work provides a scalable solution for long-term home-based monitoring of sleep apnea, enhancing the feasibility of consistent patient care.