Developing prediction algorithms for late-life depression using wearable devices: a cohort study protocol.

INTRODUCTION: Despite the high prevalence of major depressive disorder (MDD) among the elderly population, the rate of treatment is low due to stigmas and barriers to medical access. Wearable devices such as smartphones and smartwatches can help to screen MDD symptoms earlier in a natural setting while forgoing these concerns. However, previous research using wearable devices has mostly targeted the younger population. By collecting longitudinal data using wearable devices from the elderly population, this research aims to produce prediction algorithms for late-life depression and to develop strategies that strengthen medical access in community care systems. METHODS AND ANALYSIS: The current cohort study recruited a subsample of 685 elderly people from the Korean Genome and Epidemiology Study-Cardiovascular Disease Association Study, a national large cohort established in 2004. The current study has been conducted over a 3-year period to explore the development patterns of late-life depression. Participants have completed three annual face-to-face interviews (baseline, the first follow-up and the second follow-up) and 2 years of app-based surveys and passive sensing data collection. All the data collection will end at the second follow-up interview. The collected self-report, observational and passive sensing data will be primarily analysed by machine learning. ETHICS AND DISSEMINATION: This study protocol has been reviewed and approved by the Yonsei University Mirae Campus Institutional Review Board (1041849-2 02 111 SB-180-06) in South Korea. All participants provided written informed consent. The findings of this research will be disseminated by academic publications and conference presentations.

SleepMI: An AI-based screening algorithm for myocardial infarction using nocturnal electrocardiography.

Myocardial infarction (MI) is a common cardiovascular disease, the early diagnosis of which is essential for effective treatment and reduced mortality. Therefore, novel methods are required for automatic screening or early diagnosis of MI, and many studies have proposed diverse conventional methods for its detection. In this study, we aimed to develop a sleep-myocardial infarction (sleepMI) algorithm for automatic screening of MI based on nocturnal electrocardiography (ECG) findings from diagnostic polysomnography (PSG) data using artificial intelligence (AI) models. The proposed sleepMI algorithm was designed using representation and ensemble learning methods and optimized via dropout and batch normalization. In the sleepMI algorithm, a deep convolutional neural network and light gradient boost machine (LightGBM) models were mixed to obtain robust and stable performance for screening MI from nocturnal ECG findings. The nocturnal ECG signal was extracted from 2,691 participants (2,331 healthy individuals and 360 patients with MI) from the PSG data of the second follow-up stage of the Sleep Heart Health Study. The nocturnal ECG signal was extracted 3 h after sleep onset and segmented at 30-s intervals for each participant. All ECG datasets were divided into training, validation, and test sets consisting of 574,729, 143,683, and 718,412 segments, respectively. The proposed sleepMI model exhibited very high performance with precision, recall, and F1-score of 99.38%, 99.38%, and 99.38%, respectively. The total mean accuracy for automatic screening of MI using a nocturnal single-lead ECG was 99.387%. MI events can be detected using conventional 12-lead ECG signals and polysomnographic ECG recordings using our model.

Childhood adversity and late-life depression: moderated mediation model of stress and social support.

Background: As life expectancy increases, understanding the mechanism for late-life depression and finding a crucial moderator becomes more important for mental health in older adults. Childhood adversity increases the risk of clinical depression even in old age. Based on the stress sensitivity theory and stress-buffering effects, stress would be a significant mediator, while social support can be a key moderator in the mediation pathways. However, few studies have tested this moderated mediation model with a sample of older adults. This study aims to reveal the association between childhood adversity and late-life depression in older adults, taking into consideration the effects of stress and social support. Methods: This study used several path models to analyze the data from 622 elderly participants who were never diagnosed with clinical depression. Results: We found that childhood adversity increases the odds ratio of depression by approximately 20% in older adults. Path model with mediation demonstrates that stress fully mediates the pathway from childhood adversity to late-life depression. Path model with moderated mediation also illustrates that social support significantly weakens the association between childhood adversity and perceived stress. Conclusion: This study provides empirical evidence to reveal a more detailed mechanism for late-life depression. Specifically, this study identifies one crucial risk factor and one protective factor, stress and social support, respectively. This brings insight into prevention of late-life depression among those who have experienced childhood adversity.

A Prediction Algorithm for Hypoglycemia Based on Support Vector Machine Using Glucose Level and Electrocardiogram.

A prediction algorithm for hypoglycemic events is proposed using glucose levels and electrocardiogram (ECG) with support vector machine (SVM). We extracted the corrected QT interval and five heart rate variability parameters from the ECG, along with glucose level from a continuous glucose monitoring system (CGMS). This feature set is used as input to the SVM, and hypoglycemic events are predicted every 5 min using the trained SVM model for up to 30 min in advance. The proposed algorithm was developed and evaluated for nine Type-1 diabetes patients in the D1NAMO dataset. The prediction sensitivity, specificity, and accuracy values for the test set were 91.1%, 87.0%, and 89.0% (10 min before); 88.0%, 84.3%, and 86.2% (20 min before); 80.1%, 83.3%, and 81.7% (30 min before), respectively. These results show higher performance of the proposed method compared to previous studies and suggest the possibility of predicting hypoglycemia in advance.

Deep Learning for Automatic Detection of Periodic Limb Movement Disorder Based on Electrocardiogram Signals.

In this study, a deep learning model (deepPLM) is shown to automatically detect periodic limb movement syndrome (PLMS) based on electrocardiogram (ECG) signals. The designed deepPLM model consists of four 1D convolutional layers, two long short-term memory units, and a fully connected layer. The Osteoporotic Fractures in Men sleep (MrOS) study dataset was used to construct the model, including training, validating, and testing the model. A single-lead ECG signal of the polysomnographic recording was used for each of the 52 subjects (26 controls and 26 patients) in the MrOS dataset. The ECG signal was normalized and segmented (10 s duration), and it was divided into a training set (66,560 episodes), a validation set (16,640 episodes), and a test set (20,800 episodes). The performance evaluation of the deepPLM model resulted in an F1-score of 92.0%, a precision score of 90.0%, and a recall score of 93.0% for the control set, and 92.0%, 93.0%, and 90.0%, respectively, for the patient set. The results demonstrate the possibility of automatic PLMS detection in patients by using the deepPLM model based on a single-lead ECG. This could be an alternative method for PLMS screening and a helpful tool for home healthcare services for the elderly population.

Noninvasive Screening Tool for Hyperkalemia Using a Single-Lead Electrocardiogram and Deep Learning: Development and Usability Study.

BACKGROUND: Hyperkalemia monitoring is very important in patients with chronic kidney disease (CKD) in emergency medicine. Currently, blood testing is regarded as the standard way to diagnose hyperkalemia (ie, using serum potassium levels). Therefore, an alternative and noninvasive method is required for real-time monitoring of hyperkalemia in the emergency medicine department. OBJECTIVE: This study aimed to propose a novel method for noninvasive screening of hyperkalemia using a single-lead electrocardiogram (ECG) based on a deep learning model. METHODS: For this study, 2958 patients with hyperkalemia events from July 2009 to June 2019 were enrolled at 1 regional emergency center, of which 1790 were diagnosed with chronic renal failure before hyperkalemic events. Patients who did not have biochemical electrolyte tests corresponding to the original 12-lead ECG signal were excluded. We used data from 855 patients (555 patients with CKD, and 300 patients without CKD). The 12-lead ECG signal was collected at the time of the hyperkalemic event, prior to the event, and after the event for each patient. All 12-lead ECG signals were matched with an electrolyte test within 2 hours of each ECG to form a data set. We then analyzed the ECG signals with a duration of 2 seconds and a segment composed of 1400 samples. The data set was randomly divided into the training set, validation set, and test set according to the ratio of 6:2:2 percent. The proposed noninvasive screening tool used a deep learning model that can express the complex and cyclic rhythm of cardiac activity. The deep learning model consists of convolutional and pooling layers for noninvasive screening of the serum potassium level from an ECG signal. To extract an optimal single-lead ECG, we evaluated the performances of the proposed deep learning model for each lead including lead I, II, and V1-V6. RESULTS: The proposed noninvasive screening tool using a single-lead ECG shows high performances with F1 scores of 100%, 96%, and 95% for the training set, validation set, and test set, respectively. The lead II signal was shown to have the highest performance among the ECG leads. CONCLUSIONS: We developed a novel method for noninvasive screening of hyperkalemia using a single-lead ECG signal, and it can be used as a helpful tool in emergency medicine.

Diffusion of a Lifelog-Based Digital Healthcare Platform for Future Precision Medicine: Data Provision and Verification Study.

We propose a method for data provision, validation, and service expansion for the spread of a lifelog-based digital healthcare platform. The platform is an operational cloud-based platform, implemented in 2020, that has launched a tool that can validate and de-identify personal information in a data acquisition system dedicated to a center. The data acquired by the platform can be processed into products of statistical analysis and artificial intelligence (AI)-based deep learning modules. Application programming interfaces (APIs) have been developed to open data and can be linked in a programmatic manner. As a standardized policy, a series of procedures were performed from data collection to external sharing. The proposed platform collected 321.42 GB of data for 146 types of data. The reliability and consistency of the data were evaluated by an information system audit institution, with a defects ratio of approximately 0.03%. We presented definitions and examples of APIs developed in 17 functional units for data opening. In addition, the suitability of the de-identification tool was confirmed by evaluating the reduced risk of re-identification using quasi-identifiers. We presented specific methods for data verification, personal information de-identification, and service provision to ensure the sustainability of future digital healthcare platforms for precision medicine. The platform can contribute to the diffusion of the platform by linking data with external organizations and research environments in safe zones based on data reliability.

Deep Convolutional Recurrent Model for Automatic Scoring Sleep Stages Based on Single-Lead ECG Signal.

BACKGROUND: Sleep stage scoring, which is an essential step in the quantitative analysis of sleep monitoring, relies on human experts and is therefore subjective and time-consuming; thus, an easy and accurate method is needed for the automatic scoring of sleep stages. METHODS: In this study, we constructed a deep convolutional recurrent (DCR) model for the automatic scoring of sleep stages based on a raw single-lead electrocardiogram (ECG). The DCR model uses deep convolutional and recurrent neural networks to apply the complex and cyclic rhythms of human sleep. It consists of three convolutional and two recurrent layers and is optimized by dropout and batch normalization. The constructed DCR model was evaluated using multiclass classification, including five-class sleep stages (wake, N1, N2, N3, and rapid eye movement (REM)) and three-class sleep stages (wake, non-REM (NREM), and REM), using a raw single-lead ECG signal. The single-lead ECG signal was collected from 112 subjects in two groups: control (52 subjects) and sleep apnea (60 subjects). The single-lead ECG signal was preprocessed, segmented at a duration of 30 s, and divided into a training set of 89 subjects and test set of 23 subjects. RESULTS: We achieved an overall accuracy of 74.2% for five classes and 86.4% for three classes. CONCLUSIONS: These results show the DCR model's superior performance over those in the previous studies, highlighting that the model can be an alternative tool for sleep monitoring and sleep screening.

Concept and Proof of the Lifelog Bigdata Platform for Digital Healthcare and Precision Medicine on the Cloud.

PURPOSE: We propose the Lifelog Bigdata Platform as a sustainable digital healthcare system based on individual-centric lifelog datasets and describe the standardization of lifelog and clinical data in its full-cycle management system. MATERIALS AND METHODS: The Lifelog Bigdata Platform was developed by Yonsei Wonju Health System on the cloud to support digital healthcare and precision medicine. It consists of five core components: data acquisition system, de-identification of individual information, lifelog integration, analyzer, and service. We designed a gathering system into a dedicated virtual machine to save lifelog or clinical outcomes and established standard guidelines for maintaining the quality of gathering procedures. We used standard integration keys to integrate the lifelog and clinical data. Metadata were generated from the data warehouse after loading combined or fragmented data on it. We analyzed the de-identified lifelog and clinical data using the lifelog analyzer to prevent and manage acute and chronic diseases through providing results of statistics on analysis. RESULTS: The big data centers were built in four hospitals and seven companies for integrating lifelog and clinical data to develop the Lifelog Bigdata Platform. We integrated and loaded lifelog big data and clinical data for 3 years. In the first year, we uploaded 94 types of data on the platform with a total capacity of 221 GB. CONCLUSION: The Lifelog Bigdata Platform is the first to combine lifelog and clinical data. The proposed standardization guidelines can be used for future platforms to achieve a virtuous cycle structure of lifelogging big data and an industrial ecosystem.

A Prediction Model of Incident Cardiovascular Disease in Patients with Sleep-Disordered Breathing.

(1) Purpose: this study proposes a method of prediction of cardiovascular diseases (CVDs) that can develop within ten years in patients with sleep-disordered breathing (SDB). (2) Methods: For the design and evaluation of the algorithm, the Sleep Heart Health Study (SHHS) data from the 3367 participants were divided into a training set, validation set, and test set in the ratio of 5:3:2. From the data during a baseline period when patients did not have any CVD, we extracted 18 features from electrography (ECG) based on signal processing methods, 30 ECG features based on artificial intelligence (AI), ten clinical risk factors for CVD. We trained the model and evaluated it by using CVD outcomes result, monitored in follow-ups. The optimal feature vectors were selected through statistical analysis and support vector machine recursive feature elimination (SVM-RFE) of the extracted feature vectors. Features based on AI, a novel proposal from this study, showed excellent performance out of all selected feature vectors. In addition, new parameters based on AI were possibly meaningful predictors for CVD, when used in addition to the predictors for CVD that are already known. The selected features were used as inputs to the prediction model based on SVM for CVD, determining the development of CVD-free, coronary heart disease (CHD), heart failure (HF), or stroke within ten years. (3) Results: As a result, the respective recall and precision values were 82.9% and 87.5% for CVD-free; 71.9% and 63.8% for CVD; 57.2% and 55.4% for CHD; 52.6% and 40.8% for HF; 52.4% and 44.6% for stroke. The F1-score between CVD and CVD-free was 76.5%, and it was 59.1% in class four. (4) Conclusion: In conclusion, our results confirm the excellence of the prediction model for CVD in patients with SDB and verify the possibility of prediction within ten years of the CVDs that may occur in patients with SDB.

AI-Enabled Algorithm for Automatic Classification of Sleep Disorders Based on Single-Lead Electrocardiogram.

Healthy sleep is an essential physiological process for every individual to live a healthy life. Many sleep disorders both destroy the quality and decrease the duration of sleep. Thus, a convenient and accurate detection or classification method is important for screening and identifying sleep disorders. In this study, we proposed an AI-enabled algorithm for the automatic classification of sleep disorders based on a single-lead electrocardiogram (ECG). An AI-enabled algorithm-named a sleep disorder network (SDN)-was designed for automatic classification of four major sleep disorders, namely insomnia (INS), periodic leg movement (PLM), REM sleep behavior disorder (RBD), and nocturnal frontal-lobe epilepsy (NFE). The SDN was constructed using deep convolutional neural networks that can extract and analyze the complex and cyclic rhythm of sleep disorders that affect ECG patterns. The SDN consists of five layers, a 1D convolutional layer, and is optimized via dropout and batch normalization. The single-lead ECG signal was extracted from the 35 subjects with the control (CNT) and the four sleep disorder groups (seven subjects of each group) in the CAP Sleep Database. The ECG signal was pre-processed, segmented at 30 s intervals, and divided into the training, validation, and test sets consisting of 74,135, 18,534, and 23,168 segments, respectively. The constructed SDN was trained and evaluated using the CAP Sleep Database, which contains not only data on sleep disorders, but also data of the control group. The proposed SDN algorithm for the automatic classification of sleep disorders based on a single-lead ECG showed very high performances. We achieved F1 scores of 99.0%, 97.0%, 97.0%, 95.0%, and 98.0% for the CNT, INS, PLM, RBD, and NFE groups, respectively. We proposed an AI-enabled method for the automatic classification of sleep disorders based on a single-lead ECG signal. In addition, it represents the possibility of the sleep disorder classification using ECG only. The SDN can be a useful tool or an alternative screening method based on single-lead ECGs for sleep monitoring and screening.

Identification of Sleep Apnea Severity Based on Deep Learning from a Short-term Normal ECG.

BACKGROUND: This paper proposes a novel method for automatically identifying sleep apnea (SA) severity based on deep learning from a short-term normal electrocardiography (ECG) signal. METHODS: A convolutional neural network (CNN) was used as an identification model and implemented using a one-dimensional convolutional, pooling, and fully connected layer. An optimal architecture is incorporated into the CNN model for the precise identification of SA severity. A total of 144 subjects were studied. The nocturnal single-lead ECG signal was collected, and the short-term normal ECG was extracted from them. The short-term normal ECG was segmented for a duration of 30 seconds and divided into two datasets for training and evaluation. The training set consists of 82,952 segments (66,360 training set, 16,592 validation set) from 117 subjects, while the test set has 20,738 segments from 27 subjects. RESULTS: F1-score of 98.0% was obtained from the test set. Mild and moderate SA can be identified with an accuracy of 99.0%. CONCLUSION: The results showed the possibility of automatically identifying SA severity based on a short-term normal ECG signal.

Automatic Prediction of Atrial Fibrillation Based on Convolutional Neural Network Using a Short-term Normal Electrocardiogram Signal

BACKGROUND: In this study, we propose a method for automatically predicting atrial fibrillation (AF) based on convolutional neural network (CNN) using a short-term normal electrocardiogram (ECG) signal. METHODS: We designed a CNN model and optimized it by dropout and normalization. One-dimensional convolution, max-pooling, and fully-connected multiple perceptron were used to analyze the short-term normal ECG. The ECG signal was preprocessed and segmented to train and evaluate the proposed CNN model. The training and test sets consisted of the two AF and one normal dataset from the MIT-BIH database. RESULTS: The proposed CNN model for the automatic prediction of AF achieved a high performance with a sensitivity of 98.6%, a specificity of 98.7%, and an accuracy of 98.7%. CONCLUSION: The results show the possibility of automatically predicting AF based on the CNN model using a short-term normal ECG signal. The proposed CNN model for the automatic prediction of AF can be a helpful tool for the early diagnosis of AF in healthcare fields.

Multiclass classification of obstructive sleep apnea/hypopnea based on a convolutional neural network from a single-lead electrocardiogram.

OBJECTIVE: In this paper, we propose a convolutional neural network (CNN)-based deep learning architecture for multiclass classification of obstructive sleep apnea and hypopnea (OSAH) using single-lead electrocardiogram (ECG) recordings. OSAH is the most common sleep-related breathing disorder. Many subjects who suffer from OSAH remain undiagnosed; thus, early detection of OSAH is important. APPROACH: In this study, automatic classification of three classes-normal, hypopnea, and apnea-based on a CNN is performed. An optimal six-layer CNN model is trained on a training dataset (45 096 events) and evaluated on a test dataset (11 274 events). The training set (69 subjects) and test set (17 subjects) were collected from 86 subjects with length of approximately 6 h and segmented into 10 s durations. MAIN RESULTS: The proposed CNN model reaches a mean [Formula: see text]-score of 93.0 for the training dataset and 87.0 for the test dataset. SIGNIFICANCE: Thus, proposed deep learning architecture achieved a high performance for multiclass classification of OSAH using single-lead ECG recordings. The proposed method can be employed in screening of patients suspected of having OSAH.

Automated Detection of Obstructive Sleep Apnea Events from a Single-Lead Electrocardiogram Using a Convolutional Neural Network.

In this study, we propose a method for the automated detection of obstructive sleep apnea (OSA) from a single-lead electrocardiogram (ECG) using a convolutional neural network (CNN). A CNN model was designed with six optimized convolution layers including activation, pooling, and dropout layers. One-dimensional (1D) convolution, rectified linear units (ReLU), and max pooling were applied to the convolution, activation, and pooling layers, respectively. For training and evaluation of the CNN model, a single-lead ECG dataset was collected from 82 subjects with OSA and was divided into training (including data from 63 patients with 34,281 events) and testing (including data from 19 patients with 8571 events) datasets. Using this CNN model, a precision of 0.99%, a recall of 0.99%, and an F1-score of 0.99% were attained with the training dataset; these values were all 0.96% when the CNN was applied to the testing dataset. These results show that the proposed CNN model can be used to detect OSA accurately on the basis of a single-lead ECG. Ultimately, this CNN model may be used as a screening tool for those suspected to suffer from OSA.

Cardiopulmonary Coupling Analysis Using Home Sleep Monitoring System Based on Air Mattress.

This study investigates the feasibility of cardiopulmonary coupling (CPC) using home sleep monitoring system. We have designed a system to measure respiratory signals and normal-to-normal (NN) interval series in a non-contact based on air mattress. Then, CPC analysis was conducted using extracted respiratory signals and NN interval series, and six CPC parameters were extracted (VLFC, LFC, HFC, e-LFC, e-LFCNB and e-LFCBB). To evaluate the proposed method, two statistical analyses were conducted between the CPC parameters extracted by the electrocardiogram-based conventional method and the air mattress-based proposed method for five patients with obstructive sleep apnea and hypopnea (OSAH). Wilcoxon's signed rank test on the CPC parameters of the two methods indicated no significant differences (p > 0.05) and Spearman's rank correlation analysis showed high positive correlations (r ≥ 0.7, p < 0.05) between the two methods. Therefore, the proposed method has the potential for performing CPC analysis using air mattress-based system.

Obstructive Sleep Apnea Screening Using a Piezo-Electric Sensor

In this study, we propose a novel method for obstructive sleep apnea (OSA) detection using a piezo-electric sensor. OSA is a relatively common sleep disorder. However, more than 80% of OSA patients remain undiagnosed. We investigated the feasibility of OSA assessment using a single-channel physiological signal to simplify the OSA screening. We detected both snoring and heartbeat information by using a piezo-electric sensor, and snoring index (SI) and features based on pulse rate variability (PRV) analysis were extracted from the filtered piezo-electric sensor signal. A support vector machine (SVM) was used as a classifier to detect OSA events. The performance of the proposed method was evaluated on 45 patients from mild, moderate, and severe OSA groups. The method achieved a mean sensitivity, specificity, and accuracy of 72.5%, 74.2%, and 71.5%; 85.8%, 80.5%, and 80.0%; and 70.3%, 77.1%, and 71.9% for the mild, moderate, and severe groups, respectively. Finally, these results not only show the feasibility of OSA detection using a piezo-electric sensor, but also illustrate its usefulness for monitoring sleep and diagnosing OSA.

Automatic classification of apnea/hypopnea events through sleep/wake states and severity of SDB from a pulse oximeter.

This study proposes a method of automatically classifying sleep apnea/hypopnea events based on sleep states and the severity of sleep-disordered breathing (SDB) using photoplethysmogram (PPG) and oxygen saturation (SpO2) signals acquired from a pulse oximeter. The PPG was used to classify sleep state, while the severity of SDB was estimated by detecting events of SpO2 oxygen desaturation. Furthermore, we classified sleep apnea/hypopnea events by applying different categorisations according to the severity of SDB based on a support vector machine. The classification results showed sensitivity performances and positivity predictive values of 74.2% and 87.5% for apnea, 87.5% and 63.4% for hypopnea, and 92.4% and 92.8% for apnea + hypopnea, respectively. These results represent better or comparable outcomes compared to those of previous studies. In addition, our classification method reliably detected sleep apnea/hypopnea events in all patient groups without bias in particular patient groups when our algorithm was applied to a variety of patient groups. Therefore, this method has the potential to diagnose SDB more reliably and conveniently using a pulse oximeter.