A deep learning-based decision support system for diagnosis of OSAS using PTT signals

Exploring the complexity of obstructive sleep apnea: findings from machine learning on diagnosis and predictive capacity of individual factors

PURPOSE

Obstructive sleep apnoea (OSA) is a prevalent sleep disorder characterized by pharyngeal airway collapse during sleep, leading to intermittent hypoxia, intrathoracic pressure swings, and sleep fragmentation. OSA is associated with various comorbidities and risk factors, contributing to its substantial economic and social burden. Machine learning (ML) techniques offer promise in predicting OSA severity and understanding its complex pathogenesis. This study aims to compare the accuracy of different ML techniques in predicting OSA severity and identify key associated factors contributing to OSA.

METHODS

Adult patients suspected of OSA underwent clinical assessments and polysomnography. Demographic, anthropometric and clinical data were collected. Five supervised ML models (logistic regression, decision tree, random forest, extreme gradient boosting, support vector machine) were employed, optimized through grid search and cross-validation.

RESULTS

ML models exhibited varied performance across OSA severity levels. SVM demonstrated the highest accuracy for mild OSA, XGBoost for moderate OSA, and random forest for severe OSA. Logistic regression showed the highest AUC for moderate and severe OSA. Anthropometric measures, gender, and hypertension were significant predictors of OSA severity.

CONCLUSION

ML models offer valuable insights into predicting OSA severity and identifying associated factors. Our findings support the relevant potential clinical utility of ML in OSA management, although further validation and refinement are warranted.

Chin EMG Scalogram-Based Deep CNN for OSA Screening

Obstructive Sleep Apnea (OSA) is a common sleep condition characterized by frequent pauses in breathing caused by the relaxation of muscles in the upper airway during sleep. These pauses manifest in changes observed in Chin Electromyography (EMG), airflow, and oxygen saturation signals. In this paper, we propose a deep convolutional neural network (DCNN) architecture for screening OSA events and normal breathing for the OSA subjects. We utilized data from 5 OSA subjects from the American Center for Psychiatry and Neurology (ACPN) database. In this paper, we achieved a validation accuracy of 80% and a testing accuracy of 75%. Additionally, we investigated the firing pattern of motor neurons for both OSA events and non-OSA events. It was observed that for OSA subjects, the firing pattern is extremely low during OSA events, indicating muscle relaxation, while for non-OSA events, activity is high throughout the entire duration. This proposed system offers easy discrimination between OSA and non-OSA events, facilitating prompt treatment for OSA patients.

Advancements in Home-Based Devices for Detecting Obstructive Sleep Apnea: A Comprehensive Study

Obstructive Sleep Apnea (OSA) is a respiratory disorder characterized by frequent breathing pauses during sleep. The apnea-hypopnea index is a measure used to assess the severity of sleep apnea and the hourly rate of respiratory events. Despite numerous commercial devices available for apnea diagnosis and early detection, accessibility remains challenging for the general population, leading to lengthy wait times in sleep clinics. Consequently, research on monitoring and predicting OSA has surged. This comprehensive paper reviews devices, emphasizing distinctions among representative apnea devices and technologies for home detection of OSA. The collected articles are analyzed to present a clear discussion. Each article is evaluated according to diagnostic elements, the implemented automation level, and the derived level of evidence and quality rating. The findings indicate that the critical variables for monitoring sleep behavior include oxygen saturation (oximetry), body position, respiratory effort, and respiratory flow. Also, the prevalent trend is the development of level IV devices, measuring one or two signals and supported by prediction software. Noteworthy methods showcasing optimal results involve neural networks, deep learning, and regression modeling, achieving an accuracy of approximately 99%.

ECG based apnea detection by multirate processing hybrid of wavelet-empirical decomposition Hjorth features extraction and neural networks

Sleep Apnea (SA) can cause health complications including heart stroke and neurological disorders. The Polysomnography (PSG) test can detect the severity of sleep disturbance. However, it is expensive and requires a dedicated sleep laboratory and expertise to examine the patients. Therefore, it is not available to a large population in developing countries. This leads to the development of cost-effective and automated patient examination methods for the detection of sleep apnea. This study suggests an approach of using the ECG signals to categorize sleep apnea. In this work, we have devised an original technique of feature space designing by intelligently hybridizing the multirate processing, a mix of wavelet-empirical mode decomposition (W-EMD), modes-based Hjorth features extraction, and Adam-based optimized Multilayer perceptron neural network (MLPNN) for automated categorization of apnea. A publicly available ECG dataset is used for evaluating the performance of the suggested approach. Experiments are performed for four different sub-bands of the considered ECG signals. For each selected sub-band, five "Intrinsic Mode Functions" (IMFs) are extracted. Onward, three Hjorth features: complexity, activity, and mobility are mined from each IMF. In this way, four feature sets are formed based on wavelet-driven selected sub-bands. The performance of optimized MLPNN, for the apnea categorization, is compared for each feature set. Five different evaluation parameters are used to assess the performance. For the same dataset, a systematic comparison with current state-of-the-artwork has been done. Results have shown a classification accuracy of 98.12%.

Pilot study of contactless sleep apnea detection based on snore signals with hardware implementation

Objective.Sleep apnea has a high incidence and is a potentially dangerous disease, and its early detection and diagnosis are challenging. Polysomnography (PSG) is considered the best approach for sleep apnea detection, but it requires cumbersome and complicated operations. Thus, it cannot satisfy the family healthcare needs.Approach.To facilitate the initial detection of sleep apnea in the home environment, we developed a sleep apnea classification model based on snoring and hybrid neural network, and implemented the well trained model in an embedded hardware platform. We used snore signals from 32 patients at Shenzhen People's Hospital. The Mel-Fbank features were extracted from snore signals to build a sleep apnea classification model based on Bi-LSTM with attention mechanism.Main results.The proposed model classified snore signals into four types: hypopnea, normal condition, obstructive sleep apnea, and central sleep apnea, with 83.52% and 62.31% accuracies, corresponding to the subject-dependence and subject-independence validation, respectively. After pruning and model quantization, at the cost of 0.81% and 0.95% accuracy loss of the subject dependence and subject independence classification, respectively, the number of model parameters and model storage space were reduced by 32.12% and 60.37%, respectively. The model exhibited accuracies of 82.71% and 61.36% based on the subject dependence and subject independence validations, respectively. When the well trained model was successfully porting and running on an STM32 ARM-embedded platform, the model accuracy was 58.85% for the four classifications based on leave-one-subject-out validation.Significance.The proposed sleep apnea detection model can be used in home healthcare for the initial detection of sleep apnea.

Application of artificial intelligence in the diagnosis of sleep apnea

STUDY OBJECTIVES

Machine learning (ML) models have been employed in the setting of sleep disorders. This review aims to summarize the existing data about the role of ML techniques in the diagnosis, classification, and treatment of sleep-related breathing disorders.

METHODS

A systematic search in Medline, EMBASE, and Cochrane databases through January 2022 was performed.

RESULTS

Our search strategy revealed 132 studies that were included in the systematic review. Existing data show that ML models have been successfully used for diagnostic purposes. Specifically, ML models showed good performance in diagnosing sleep apnea using easily obtained features from the electrocardiogram, pulse oximetry, and sound signals. Similarly, ML showed good performance for the classification of sleep apnea into obstructive and central categories, as well as predicting apnea severity. Existing data show promising results for the ML-based guided treatment of sleep apnea. Specifically, the prediction of outcomes following surgical treatment and optimization of continuous positive airway pressure therapy can be guided by ML models.

CONCLUSIONS

The adoption and implementation of ML in the field of sleep-related breathing disorders is promising. Advancements in wearable sensor technology and ML models can help clinicians predict, diagnose, and classify sleep apnea more accurately and efficiently.

CITATION

Bazoukis G, Bollepalli SC, Chung CT, et al. Application of artificial intelligence in the diagnosis of sleep apnea. J Clin Sleep Med. 2023;19(7):1337-1363.

ERS technical standards for using type III devices (limited channel studies) in the diagnosis of sleep disordered breathing in adults and children

For more than three decades, type III devices have been used in the diagnosis of sleep disordered breathing in supervised as well as unsupervised settings. They have satisfactory positive and negative predictive values for detecting obstructive and central sleep apnoea in populations with moderately high pre-test probability of symptoms associated with these events. However, standardisation of commercially available type III devices has never been undertaken and the technical specifications can vary widely. None have been subjected to the same rigorous processes as most other diagnostic modalities in the medical field. Although type III devices do not include acquisition of electroencephalographic signals overnight, the minimum number of physical sensors required to allow for respiratory event scoring using standards outlined by the American Academy of Sleep Medicine remains debatable. This technical standard summarises data on type III studies published since 2007 from multiple perspectives in both adult and paediatric sleep practice. Most importantly, it aims to provide a framework for considering current type III device limitations in the diagnosis of sleep disordered breathing while raising research- and practice-related questions aimed at improving our use of these devices in the present and future.

Physiological Status Prediction Based on a Novel Hybrid Intelligent Scheme

Physiological status plays an important role in clinical diagnosis. However, the temporal physiological data change dynamically with time, and the amount of data is large; furthermore, obtaining a complete history of data has become difficult. We propose a hybrid intelligent scheme for physiological status prediction, which can be effectively utilized to predict the physiological status of patients and provide a reference for clinical diagnosis. Our proposed scheme initially extracted the attribute information of nonlinear dynamic changes in physiological signals. The maximum discriminant feature subset was selected by employing conditional relevance mutual information feature selection. An optimal subset of features was fed into the particle swarm optimization-support vector machine classifier to perform classification. For the prediction task, the proposed hybrid intelligent scheme was tested on the Sleep Heart Health Study dataset for sleep status prediction. Experimental results demonstrate that our proposed intelligent scheme outperforms the conventional machine learning classification methods.

A review of automated sleep disorder detection

Automated sleep disorder detection is challenging because physiological symptoms can vary widely. These variations make it difficult to create effective sleep disorder detection models which support hu-man experts during diagnosis and treatment monitoring. From 2010 to 2021, authors of 95 scientific papers have taken up the challenge of automating sleep disorder detection. This paper provides an expert review of this work. We investigated whether digital technology and Artificial Intelligence (AI) can provide automated diagnosis support for sleep disorders. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines during the content discovery phase. We compared the performance of proposed sleep disorder detection methods, involving differ-ent datasets or signals. During the review, we found eight sleep disorders, of which sleep apnea and insomnia were the most studied. These disorders can be diagnosed using several kinds of biomedical signals, such as Electrocardiogram (ECG), Polysomnography (PSG), Electroencephalogram (EEG), Electromyogram (EMG), and snore sound. Subsequently, we established areas of commonality and distinctiveness. Common to all reviewed papers was that AI models were trained and tested with labelled physiological signals. Looking deeper, we discovered that 24 distinct algorithms were used for the detection task. The nature of these algorithms evolved, before 2017 only traditional Machine Learning (ML) was used. From 2018 onward, both ML and Deep Learning (DL) methods were used for sleep disorder detection. The strong emergence of DL algorithms has considerable implications for future detection systems because these algorithms demand significantly more data for training and testing when compared with ML. Based on our review results, we suggest that both type and amount of labelled data is crucial for the design of future sleep disorder detection systems because this will steer the choice of AI algorithm which establishes the desired decision support. As a guiding principle, more labelled data will help to represent the variations in symptoms. DL algorithms can extract information from these larger data quantities more effectively, therefore; we predict that the role of these algorithms will continue to expand.

Automatic diagnosis of sleep apnea from biomedical signals using artificial intelligence techniques: Methods, challenges, and future works

Apnea is a sleep disorder that stops or reduces airflow for a short time during sleep. Sleep apnea may last for a few seconds and happen for many while sleeping. This reduction in breathing is associated with loud snoring, which may awaken the person with a feeling of suffocation. So far, a variety of methods have been introduced by researchers to diagnose sleep apnea, among which the polysomnography (PSG) method is known to be the best. Analysis of PSG signals is very complicated. Many studies have been conducted on the automatic diagnosis of sleep apnea from biological signals using artificial intelligence (AI), including machine learning (ML) and deep learning (DL) methods. This research reviews and investigates the studies on the diagnosis of sleep apnea using AI methods. First, computer aided diagnosis system (CADS) for sleep apnea using ML and DL techniques along with its parts including dataset, preprocessing, and ML and DL methods are introduced. This research also summarizes the important specifications of the studies on the diagnosis of sleep apnea using ML and DL methods in a table. In the following, a comprehensive discussion is made on the studies carried out in this field. The challenges in the diagnosis of sleep apnea using AI methods are of paramount importance for researchers. Accordingly, these obstacles are elaborately addressed. In another section, the most important future works for studies on sleep apnea detection from PSG signals and AI techniques are presented. Ultimately, the essential findings of this study are provided in the conclusion section.

This article is categorized under:

Diagnosis of obstructive sleep apnea with prediction of flow characteristics according to airway morphology automatically extracted from medical images: Computational fluid dynamics and artificial intelligence approach

BACKGROUND

Obstructive sleep apnea syndrome (OSAS) is being observed in an increasing number of cases. It can be diagnosed using several methods such as polysomnography.

OBJECTIVES

To overcome the challenges of time and cost faced by conventional diagnostic methods, this paper proposes computational fluid dynamics (CFD) and machine-learning approaches that are derived from the upper-airway morphology with automatic segmentation using deep learning.

METHOD

We adopted a 3D UNet deep-learning model to perform medical image segmentation. 3D UNet prevents the feature-extraction loss that may occur by concatenating layers and extracts the anteroposterior coordination and width of the airway morphology. To create flow characteristics of the upper airway training data, we analyzed the changes in flow characteristics according to the upper-airway morphology using CFD. A multivariate Gaussian process regression (MVGPR) model was used to train the flow characteristic values. The trained MVGPR enables the prompt prediction of the aerodynamic features of the upper airway without simulation. Unlike conventional regression methods, MVGPR can be trained by considering the correlation between the flow characteristics. As a diagnostic step, a support vector machine (SVM) with predicted aerodynamic and biometric features was used in this study to classify patients as healthy or suffering from moderate OSAS. SVM is beneficial as it is easy to learn even with a small dataset, and it can diagnose various flow characteristics as factors while enhancing the feature via the kernel function. As the patient dataset is small, the Monte Carlo cross-validation was used to validate the trained model. Furthermore, to overcome the imbalanced data problem, the oversampling method was applied.

RESULT

The segmented upper-airway results of the high-resolution and low-resolution models present overall average dice coefficients of 0.76±0.041 and 0.74±0.052, respectively. Furthermore, the classification accuracy, sensitivity, specificity, and F1-score of the diagnosis algorithm were 81.5%, 89.3%, 86.2%, and 87.6%, respectively.

CONCLUSION

The convenience and accuracy of sleep apnea diagnosis are improved using deep learning and machine learning. Further, the proposed method can aid clinicians in making appropriate decisions to evaluate the possible applications of OSAS.

AIOSA: An approach to the automatic identification of obstructive sleep apnea events based on deep learning

Obstructive Sleep Apnea Syndrome (OSAS) is the most common sleep-related breathing disorder. It is caused by an increased upper airway resistance during sleep, which determines episodes of partial or complete interruption of airflow. The detection and treatment of OSAS is particularly important in patients who suffered a stroke, because the presence of severe OSAS is associated with higher mortality, worse neurological deficits, worse functional outcome after rehabilitation, and a higher likelihood of uncontrolled hypertension. The gold standard test for diagnosing OSAS is polysomnography (PSG). Unfortunately, performing a PSG in an electrically hostile environment, like a stroke unit, on neurologically impaired patients is a difficult task; moreover, the number of strokes per day vastly outnumbers the availability of polysomnographs and dedicated healthcare professionals. Hence, a simple and automated recognition system to identify OSAS cases among acute stroke patients, relying on routinely recorded vital signs, is highly desirable. The vast majority of the work done so far focuses on data recorded in ideal conditions and highly selected patients, and thus it is hardly exploitable in real-life circumstances, where it would be of actual use. In this paper, we propose a novel convolutional deep learning architecture able to effectively reduce the temporal resolution of raw waveform data, like physiological signals, extracting key features that can be used for further processing. We exploit models based on such an architecture to detect OSAS events in stroke unit recordings obtained from the monitoring of unselected patients. Unlike existing approaches, annotations are performed at one-second granularity, allowing physicians to better interpret the model outcome. Results are considered to be satisfactory by the domain experts. Moreover, through tests run on a widely-used public OSAS dataset, we show that the proposed approach outperforms current state-of-the-art solutions.

Prediction of obstructive sleep apnea using ensemble of recurrence plot convolutional neural networks (RPCNNs) from polysomnography signals

Obstructive Sleep Apnea (OSA) is a common disorder characterized by periodic cessation of breathing during sleep. OSA affects daily life and poses a severe threat to human health. The standard clinical method for identifying and predicting OSA events is the use of Polysomnography signals. In this paper, a novel scheme based on an ensemble of recurrence plots (RPs) and pre-trained convolutional neural networks (RPCNNs) is proposed to improve the prediction rate of OSA. First, RPs were used to represent the dynamic behavior of single electroencephalogram (EEG) and electrocardiogram (ECG) signals for 60 s before and during OSA events. Then, using RPs, three prompt CNNs named ResNet-50 were fine-tuned, and their classification results were fused via the Majority Voting (MV) method to produce a final result concerning prediction. Next, the subject-independent Leave-One-Subject-Out Cross-Validation (LOSO-CV) and subject-dependent 10-fold Cross-Validation (10-fold CV) methods were used to validate the prediction rate from signals derived from the University College Dublin Sleep Apnea Database. Finally, the highest achieved average accuracy for the fusion level was 91.74% and 89.45% at the 10-fold CV and LOSO-CV. Additionally, our results outperformed state-of-the-art findings and could be recommended to predict and detect other biomedical signals. As a result, this predictive system can also be used to adjust the air pressure in sleep apnea patients' Automatic Positive Airway Pressure (APAP) devices.

Obstructive sleep apnea prediction from electrocardiogram scalograms and spectrograms using convolutional neural networks

Objective.In this study, we conducted a comparative analysis of deep convolutional neural network (CNN) models in predicting obstructive sleep apnea (OSA) using electrocardiograms. Unlike other studies in the literature, this study automatically extracts time-frequency features by using CNNs instead of manual feature extraction from ECG recordings.Approach.The proposed model generates scalogram and spectrogram representations by transforming preprocessed 30 s ECG segments from time domain to the frequency domain using continuous wavelet transform and short time Fourier transform, respectively. We examined AlexNet, GoogleNet and ResNet18 models in predicting OSA events. The effect of transfer learning on success is also investigated. Based on the observed results, we proposed a new model that is found more effective in estimation. In total, 152 ECG recordings were included in the study for training and evaluation of the models.Main results.The prediction using scalograms immediately 30 s before potential OSA onsets gave the best performance with 82.30% accuracy, 83.22% sensitivity, 82.27% specificity and 82.95% positive predictive value. The prediction using spectrograms also achieved up to 80.13% accuracy and 81.99% sensitivity on prediction. Per-recording classification suggested considerable results with 91.93% accuracy for prediction of OSA events.Significance.Time-frequency deep features of scalograms and spectrograms of ECG segments prior to OSA events provided reliable information about the possible events in the future. The proposed CNN model can be used as a good indicator to accurately predict OSA events using ECG recordings.

COVIDetectioNet: COVID-19 diagnosis system based on X-ray images using features selected from pre-learned deep features ensemble

The recent novel coronavirus (also known as COVID-19) has rapidly spread worldwide, causing an infectious respiratory disease that has killed hundreds of thousands and infected millions. While test kits are used for diagnosis of the disease, the process takes time and the test kits are limited in their availability. However, the COVID-19 disease is also diagnosable using radiological images taken through lung X-rays. This process is known to be both faster and more reliable as a form of identification and diagnosis. In this regard, the current study proposes an expert-designed system called COVIDetectioNet model, which utilizes features selected from combination of deep features for diagnosis of COVID-19. For this purpose, a pretrained Convolutional Neural Network (CNN)-based AlexNet architecture that employed the transfer learning approach, was used. The effective features that were selected using the Relief feature selection algorithm from all layers of the architecture were then classified using the Support Vector Machine (SVM) method. To verify the validity of the model proposed, a total of 6092 X-ray images, classified as Normal (healthy), COVID-19, and Pneumonia, were obtained from a combination of public datasets. In the experimental results, an accuracy of 99.18% was achieved using the model proposed. The results demonstrate that the proposed COVIDetectioNet model achieved a superior level of success when compared to previous studies.

Discrimination of β-thalassemia and iron deficiency anemia through extreme learning machine and regularized extreme learning machine based decision support system

The symptoms of Iron Deficiency Anemia (IDA) and β-thalassemia (β-TT) disease are similar and the distinction between them is time consuming and costly. There are several indices used to differentiate IDA from β-thalassemia disease. Complete Blood Count (CBC) is a rapid, inexpensive and accessible test for the diagnosis of anemia and is used as a primary test. However, since CBC cannot fully distinguish between IDA and β-thalassemia, more advanced testing is required. These tests are not available in small centers and are performed on higher-cost devices. Moreover, it is important to differentiate between anemia and β-thalassemia medically for two reasons (IDA). First, if a patient with β-Thalassemia is diagnosed with IDA, the patient is given unnecessary iron supplementation as a result of the treatment, which is recommended by the doctor. Secondly, when the patient with β-thalassemia is diagnosed with IDA, children will have β-thalassemia patients in marriages. A decision support system to distinguish between β-Thalassemia and IDA has been developed. Logistic Regression, K-Nearest Neighbours, Support Vector Machine, Extreme Learning Machine and Regularized Extreme Learning Machine classification algorithms were used in the proposed system. Classification performance was evaluated with Accuracy, sensitivity, f-measure, Specificty parameters using Hemoglobin, RBC, HCT, MCV, MCH, MCHC and RDW parameters obtained from 342 patients. 96.30% accuracy for female, 94.37% for male, and 95.59% in co-evaluation of male and female patients were obtained.

Multi-label classification of retinal lesions in diabetic retinopathy for automatic analysis of fundus fluorescein angiography based on deep learning

PURPOSE

To automatically detect and classify the lesions of diabetic retinopathy (DR) in fundus fluorescein angiography (FFA) images using deep learning algorithm through comparing 3 convolutional neural networks (CNNs).

METHODS

A total of 4067 FFA images from Eye Center at the Second Affiliated Hospital of Zhejiang University School of Medicine were annotated with 4 kinds of lesions of DR, including non-perfusion regions (NP), microaneurysms, leakages, and laser scars. Three CNNs including DenseNet, ResNet50, and VGG16 were trained to achieve multi-label classification, which means the algorithms could identify 4 retinal lesions above at the same time.

RESULTS

The area under the curve (AUC) of DenseNet reached 0.8703, 0.9435, 0.9647, and 0.9653 for detecting NP, microaneurysms, leakages, and laser scars, respectively. For ResNet50, AUC was 0.8140 for NP, 0.9097 for microaneurysms, 0.9585 for leakages, and 0.9115 for laser scars. And for VGG16, AUC was 0.7125 for NP, 0.5569 for microaneurysms, 0.9177 for leakages, and 0.8537 for laser scars.

CONCLUSIONS

Experimental results demonstrate that DenseNet is a suitable model to automatically detect and distinguish retinal lesions in the FFA images with multi-label classification, which lies the foundation of automatic analysis for FFA images and comprehensive diagnosis and treatment decision-making for DR.

Is it possible to detect cerebral dominance via EEG signals by using deep learning?

Each brain hemisphere is dominant for certain functions such as speech. The determination of speech laterality prior to surgery is of paramount importance for accurate risk prediction. In this study, we aimed to determine speech laterality via EEG signals by using noninvasive machine learning techniques. The retrospective study included 67 subjects aged 18-65 years who had no chronic diseases and were diagnosed as healthy based on EEG examination. The subjects comprised 35 right-hand dominant (speech center located in the left hemisphere) and 32 left-hand dominant individuals (speech center located in the right hemisphere). A spectrogram was created for each of the 18 EEG channels by using various Convolutional Neural Networks (CNN) architectures including VGG16, VGG19, ResNet, MobileNet, NasNet, and DenseNet. These architectures were used to extract features from the spectrograms. The extracted features were classified using Support Vector Machines (SVM) and the classification performances of the CNN models were evaluated using Area Under the Curve (AUC). Of all the CNN models used in the study, VGG16 had a higher AUC value (0.83 ± 0.05) in the determination of speech laterality compared to all other models. The present study is a pioneer investigation into the determination of speech laterality via EEG signals with machine learning techniques, which, to our knowledge, has never been reported in the literature. Moreover, the classification results obtained in the study are promising and lead the way for subsequent studies though not practically feasible.