Over the years, social determinants of health have increasingly been discovered to have a significant impact on an extensive range of mental and physical health outcomes. The wide adoption of electronic health records has made it possible for the development of automated techniques to conduct studies on the effects of such factors on health. With the recent advancements in machine learning, it has become the key technology for extracting patient-level social and behavioral factors from electronic health records. However, the current state-of-the-art machine learning techniques used to extract social and behavioral factors have yet to be deciphered and compared. This narrative review aimed at evaluating advancements in machine learning technology used in the literature over the past decade for the extraction of social determinants of health factors from electronic health records data to gain a better understanding of how and when to leverage them. This was conducted by analyzing the social determinants of health categories, evaluating the relationship between these characteristics and patient health outcomes, reviewing documentation practices of the characteristics in electronic health records, summarizing and comparing current machine learning techniques in the field as well as assessing their limitation, and suggesting some future directions of study. Leveraging machine learning can overcome challenges faced with parsing unstructured clinical data, ease the extraction of relevant concepts in medical text, and aid in studying the health risks caused by social behavior. Despite current knowledge of the significant impact of social and behavioral factors on health, they are rarely documented and investigated in healthcare systems. Understanding current state-of-the-art machine learning technology to support the identification of social determinants of health in a clinical setting can influence clinical decision-making to provide better overall patient wellness with the ultimate goal of producing health equity.

Dysphagia occurs secondary to a variety of underlying etiologies and can contribute to increased risk of adverse events such as aspiration pneumonia and premature mortality. Dysphagia is primarily diagnosed and characterized by instrumental swallowing exams such as videofluoroscopic swallowing studies. videofluoroscopic swallowing studies involve the inspection of a series of radiographic images for signs of swallowing dysfunction. Though effective, videofluoroscopic swallowing studies are only available in certain clinical settings and are not always desirable or feasible for certain patients. Because of the limitations of current instrumental swallow exams, research studies have explored the use of acceleration signals collected from neck sensors and demonstrated their potential in providing comparable radiation-free diagnostic value as videofluoroscopic swallowing studies. In this study, we used a hybrid deep convolutional recurrent neural network that can perform multi-level feature extraction (localized and across time) to annotate swallow segments automatically via multi-channel swallowing acceleration signals. In total, we used signals and videofluoroscopic swallowing study images of 3144 swallows from 248 patients with suspected dysphagia. Compared to other deep network variants, our network was superior at detecting swallow segments with an average area under the receiver operating characteristic curve value of 0.82 (95% confidence interval: 0.807-0.841), and was in agreement with up to 90% of the gold standard-labeled segments.

Artificial intelligence and machine learning techniques have progressed dramatically and become powerful tools required to solve complicated tasks, such as computer vision, speech recognition, and natural language processing. Since these techniques have provided promising and evident results in these fields, they emerged as valuable methods for applications in human physiology and healthcare. General physiological recordings are time-related expressions of bodily processes associated with health or morbidity. Sequence classification, anomaly detection, decision making, and future status prediction drive the learning algorithms to focus on the temporal pattern and model the nonstationary dynamics of the human body. These practical requirements give birth to the use of recurrent neural networks (RNNs), which offer a tractable solution in dealing with physiological time series and provide a way to understand complex time variations and dependencies. The primary objective of this article is to provide an overview of current applications of RNNs in the area of human physiology for automated prediction and diagnosis within different fields. Finally, we highlight some pathways of future RNN developments for human physiology.

As different scientific disciplines begin to converge on machine learning for causal inference, we demonstrate the application of machine learning algorithms in the context of longitudinal causal estimation using electronic health records. Our aim is to formulate a marginal structural model for estimating diabetes care provisions in which we envisioned hypothetical (i.e. counterfactual) dynamic treatment regimes using a combination of drug therapies to manage diabetes: metformin, sulfonylurea and SGLT-2i. The binary outcome of diabetes care provisions was defined using a composite measure of chronic disease prevention and screening elements [27] including (i) primary care visit, (ii) blood pressure, (iii) weight, (iv) hemoglobin A1c, (v) lipid, (vi) ACR, (vii) eGFR and (viii) statin medication. We used several statistical learning algorithms to describe causal relationships between the prescription of three common classes of diabetes medications and quality of diabetes care using the electronic health records contained in National Diabetes Repository. In particular, we generated an ensemble of statistical learning algorithms using the SuperLearner framework based on the following base learners: (i) least absolute shrinkage and selection operator, (ii) ridge regression, (iii) elastic net, (iv) random forest, (v) gradient boosting machines, and (vi) neural network. Each statistical learning algorithm was fitted using the pseudo-population generated from the marginalization of the time-dependent confounding process. Covariate balance was assessed using the longitudinal (i.e. cumulative-time product) stabilized weights with calibrated restrictions. Our results indicated that the treatment drop-in cohorts (with respect to metformin, sulfonylurea and SGLT-2i) may have improved diabetes care provisions in relation to treatment naïve (i.e. no treatment) cohort. As a clinical utility, we hope that this article will facilitate discussions around the prevention of adverse chronic outcomes associated with type II diabetes through the improvement of diabetes care provisions in primary care.

Aspiration is a serious complication of swallowing disorders. Adequate detection of aspiration is essential in dysphagia management and treatment. High-resolution cervical auscultation has been increasingly considered as a promising noninvasive swallowing screening tool and has inspired automatic diagnosis with advanced algorithms. The performance of such algorithms relies heavily on the amount of training data. However, the practical collection of cervical auscultation signal is an expensive and time-consuming process because of the clinical settings and trained experts needed for acquisition and interpretations. Furthermore, the relatively infrequent incidence of severe airway invasion during swallowing studies constrains the performance of machine learning models. Here, we produced supplementary training exemplars for desired class by capturing the underlying distribution of original cervical auscultation signal features using auxiliary classifier Wasserstein generative adversarial networks. A 10-fold subject cross-validation was conducted on 2079 sets of 36-dimensional signal features collected from 189 patients undergoing swallowing examinations. The proposed data augmentation outperforms basic data sampling, cost-sensitive learning and other generative models with significant enhancement. This demonstrates the remarkable potential of proposed network in improving classification performance using cervical auscultation signals and paves the way of developing accurate noninvasive swallowing evaluation in dysphagia care.

Upper esophageal sphincter is an important anatomical landmark of the swallowing process commonly observed through the kinematic analysis of radiographic examinations that are vulnerable to subjectivity and clinical feasibility issues. Acting as the doorway of esophagus, upper esophageal sphincter allows the transition of ingested materials from pharyngeal into esophageal stages of swallowing and a reduced duration of opening can lead to penetration/aspiration and/or pharyngeal residue. Therefore, in this study we consider a non-invasive high resolution cervical auscultation-based screening tool to approximate the human ratings of upper esophageal sphincter opening and closure. Swallows were collected from 116 patients and a deep neural network was trained to produce a mask that demarcates the duration of upper esophageal sphincter opening. The proposed method achieved more than 90% accuracy and similar values of sensitivity and specificity when compared to human ratings even when tested over swallows from an independent clinical experiment. Moreover, the predicted opening and closure moments surprisingly fell within an inter-human comparable error of their human rated counterparts which demonstrates the clinical significance of high resolution cervical auscultation in replacing ionizing radiation-based evaluation of swallowing kinematics.

In this paper, we introduce a transfer learning approach for our novel hybrid brain-computer interface in which electroencephalography and functional transcranial Doppler ultrasound are used simultaneously to record brain electrical activity and cerebral blood velocity respectively due to flickering mental rotation and word generation tasks. We reduced each trial into a scalar score using Regularized Discriminant Analysis (RDA). For each individual, class conditional probabilistic distribution of each mental task was estimated using RDA scores of the trials corresponding to that mental task. Similarities between class conditional distributions across individuals were measured using Kullback-Leibler divergence, Bhattacharyya, and Hellinger distances. Classification task was performed using Quadratic Discriminant Analysis (QDA), Linear Discriminant Analysis (LDA), and Support Vector Machines (SVM). We demonstrate that transfer learning can reduce calibration requirements up to %87.5. Moreover, it was found that QDA provides the most significant performance improvement compared to the case when no transfer learning is employed.

The Internet of Medical Things (IoMT) designates the interconnection of communication-enabled medical-grade devices and their integration to wider-scale health networks in order to improve patients' health. However, because of the critical nature of health-related systems, the IoMT still faces numerous challenges, more particularly in terms of reliability, safety, and security. In this paper, we present a comprehensive literature review of recent contributions focused on improving the IoMT through the use of formal methodologies provided by the cyber-physical systems community. We describe the practical application of the democratization of medical devices for both patients and health-care providers. We also identify unexplored research directions and potential trends to solve uncharted research problems.

This paper presents the development and real-time testing of an automated expert diagnostic telehealth system for the diagnosis of 2 respiratory diseases, asthma and Chronic Obstructive Pulmonary Disease (COPD). The system utilizes Android, Java, MATLAB, and PHP technologies and consists of a spirometer, mobile application, and expert diagnostic system. To evaluate the effectiveness of the system, a prospective study was carried out in 3 remote primary healthcare institutions, and one hospital in Bosnia and Herzegovina healthcare system. During 6 months, 780 patients were assessed and diagnosed with an accuracy of 97.32%. The presented approach is simple to use and offers specialized consultations for patients in remote, rural, and isolated communities, as well as old and less physically mobile patients. While improving the quality of care delivered to patients, it was also found to be very beneficial in terms of healthcare.

Vertex-frequency analysis of graph signals is a challenging topic for research and applications. Counterparts of the short-time Fourier transform, the wavelet transform, and the Rihaczek distribution have recently been introduced to the graph-signal analysis. In this letter, we have extended the energy distributions to a general reduced interference distributions class. It can improve the vertex-frequency representation of a graph signal while preserving the marginal properties. This class is related to the spectrogram of graph signals as well. Efficiency of the proposed representations is illustrated in examples.

Deep learning, a relatively new branch of machine learning, has been investigated for use in a variety of biomedical applications. Deep learning algorithms have been used to analyze different physiological signals and gain a better understanding of human physiology for automated diagnosis of abnormal conditions. In this paper, we provide an overview of deep learning approaches with a focus on deep belief networks in electroencephalography applications. We investigate the state-of-the-art algorithms for deep belief networks and then cover the application of these algorithms and their performances in electroencephalographic applications. We covered various applications of electroencephalography in medicine, including emotion recognition, sleep stage classification, and seizure detection, in order to understand how deep learning algorithms could be modified to better suit the tasks desired. This review is intended to provide researchers with a broad overview of the currently existing deep belief network methodology for electroencephalography signals, as well as to highlight potential challenges for future research.

In this paper, the use of mutual information and the Learn++.NSE algorithm is proposed to create an EEG SSVEP BCI system that can select and utilize data sets originating from a group of users. In typical BCI systems, the nonstationarity in the EEG prevents the system from blindly applying training data from other users to the incoming data. Mutual information is introduced to select previous data sets that provide the most information about current random variables. A signed rank test was employed to show that this configuration outperformed both normal Learn++.NSE ensembles and LDA classifiers. This indicates that mutual information and ensemble learning techniques may prove useful in improving user transferability in SSVEP systems with low computational requirements.

Wireless transcutaneous power transfer and communication has the potential to reduce the size of implantable medical devices, thereby reducing patient discomfort and minimizing the tissue area exposed to foreign material. Electromagnetic transmission mechanisms through tissue are determined by tissue structure and associated frequency-dependent tissue properties, which are significant in the design of wireless implantable medical devices. The purpose of this study was to investigate the effects of varying tissue dielectric properties on maximum power transfer to a subcutaneously implanted device in a paired electrode system designed for use in proximity to metallic orthopedic implants. The transcutaneous system including external and implanted electrode pairs was simulated at several radio frequencies (125 kHz, 1 MHz, 13.56 MHz, 403 MHz, and 915 MHz) while varying the dielectric properties of the tissue medium over a range of physiological values. Maximum power transfer was calculated to represent the best-case power gain across the range of tissue properties and frequencies, and greater achievable efficiencies were seen with higher quality factor as a function of the tissue properties. The results suggest that in the paired electrode system, utilization of capacitive coupling allows the system to function in proximity to metallic surfaces such as orthopedic implants. The results also suggest that higher power gains are possible through a choice of implant location based on expected tissue properties.

Cervical auscultation is the recording of sounds and vibrations caused by the human body from the throat during swallowing. While traditionally done by a trained clinician with a stethoscope, much work has been put towards developing more sensitive and clinically useful methods to characterize the data obtained with this technique. The eventual goal of the field is to improve the effectiveness of screening algorithms designed to predict the risk that swallowing disorders pose to individual patients' health and safety. This paper provides an overview of these signal processing techniques and summarizes recent advances made with digital transducers in hopes of organizing the highly varied research on cervical auscultation. It investigates where on the body these transducers are placed in order to record a signal as well as the collection of analog and digital filtering techniques used to further improve the signal quality. It also presents the wide array of methods and features used to characterize these signals, ranging from simply counting the number of swallows that occur over a period of time to calculating various descriptive features in the time, frequency, and phase space domains. Finally, this paper presents the algorithms that have been used to classify this data into 'normal' and 'abnormal' categories. Both linear as well as non-linear techniques are presented in this regard.

Unlike synchronous processing, asynchronous processing is more efficient in biomedical and sensing networks applications as it is free from aliasing constraints and quantization error in the amplitude, it allows continuous–time processing and more importantly data is only acquired in significant parts of the signal. We consider signal decomposers based on the asynchronous sigma delta modulator (ASDM), a non‐linear feedback system that maps the signal amplitude into the zero‐crossings of a binary output signal. The input, the zero‐crossings and the ASDM parameters are related by an integral equation making the signal reconstruction difficult to implement. Modifying the model for the ASDM, we obtain a recursive equation that permits to obtain the non‐uniform samples from the zero‐time crossing values. Latticing the joint time‐frequency space into defined frequency bands, and time windows depending on the scale parameter different decompositions are possible. We present two cascade low‐ and high‐frequency decomposers, and a bank‐of‐filters parallel decomposer. This last decomposer using the modified ASDM behaves like a asynchronous analog to digital converter, and using an interpolator based on Prolate Spheroidal Wave functions allows reconstruction of the original signal. The asynchronous approaches proposed here are well suited for processing signals sparse in time, and for low‐power applications.

An implantable wireless Doppler device used in microsurgical free flap surgeries can suffer from lost data points. To recover the lost samples, the authors considered the approaches based on a recently proposed compressive sensing. In this paper, they performed a comparative analysis of several different approaches by using synthetic and real signals obtained during blood flow monitoring in four pigs. They considered three different bases functions: Fourier bases, discrete prolate spheroidal sequences and modulated discrete prolate spheroidal sequences, respectively. To avoid the computational burden, they considered the approaches based on the l 1 minimisation for all the three bases. To understand the trade‐off between the computational complexity and the accuracy, they also used a recovery process based on a matching pursuit and modulated discrete prolate spheroidal sequences bases. For both the synthetic and the real signals, the matching approach with modulated discrete prolate spheroidal sequences provided the most accurate results. Future studies should focus on the optimisation of the modulated discrete prolate spheroidal sequences in order to further decrease the computational complexity and increase the accuracy.

We present a novel approach to estimating the mean square error (MSE) associated with any given threshold level in both hard and soft thresholding. The estimate is provided by using only the data that is being thresholded. This adaptive approach provides probabilistic confidence bounds on the MSE. The MSE bounds can be used to evaluate the denoising method. Our simulation results confirm that not only does the method provide an accurate estimate of the MSE for any given thresohlding method, but the proposed method can also search and find an optimum threshold for any noisy data with regard to MSE.

The reassignment method is a widespread approach for obtaining high resolution time-frequency representations. Nevertheless, its performance is not always optimal and can deteriorate for low signal-to-noise ratio (SNR) values. In order to overcome these obstacles, a novel method for obtaining high resolution time-frequency representations is proposed in this paper. The new method implements proposed nonparametric snakes in order to obtain accurate locations of the signal ridges in the time-frequency domain. The results of numerical analysis show that the proposed method is capable of achieving significantly higher concentration of signals in the time-frequency domain in comparison to the spectrogram and the traditional reassignment method. Furthermore, the new scheme also maintains good performance for low SNR values, while the performance of the other two considered methods significantly diminishes. It is clear from the results that the proposed method might be of significance in applications where accurate estimation of the signal components is required for low SNR values.

Accurate and sparse representation of a moderately fast fading channel using bases functions is achievable when both channel and bases bands align. If a mismatch exists, usually a larger number of bases functions is needed to achieve the same accuracy. In this paper, we propose a novel approach for channel estimation based on frames, which preserves sparsity and improves estimation accuracy. Members of the frame are formed by modulating and varying the band-width of discrete prolate spheroidal sequences (DPSS) in order to reflect various scattering scenarios. To achieve the sparsity of the proposed representation, a matching pursuit approach is employed. The estimation accuracy of the scheme is evaluated and compared with the accuracy of a Slepian basis expansion estimator based on DPSS for a variety of mobile channel parameters. The results clearly indicate that for the same number of atoms, a significantly higher estimation accuracy is achievable with the proposed scheme when compared to the DPSS estimator.

Instantaneous frequency (IF) is a fundamental concept that can be found in many disciplines such as communications, speech, and music processing. In this letter, analysis of an IF estimator, based on a time-frequency technique known as S-transform, is performed. The performance analysis is carried out in a white Gaussian noise environment, and expressions for the bias and the variance of the estimator are determined. The results show that the bias and the variance are signal dependent. This has been statistically confirmed through numerical simulations of several signal classes.

A time-frequency signal analysis tool, known as S-transform, can suffer from poor energy concentration in the time-frequency domain. In this paper, a frequency dependent Kaiser window is presented for improving the energy concentration of the S-transform. The new window is analyzed using a set of test signals. The results indicate that the proposed scheme can significantly improve the energy concentration in the time-frequency domain in comparison with the standard S-transform.

In this paper, a novel correlation-based pattern classifier that relies on the analysis of time–frequency decomposition of a template and signals is proposed. Significant improvements in resolution and accuracy are obtained using this new classifier when compared to a conventional correlation-based one. The short-time Fourier transform, continuous wavelet transform, and S-transform are considered in the time–frequency decomposition process. To evaluate the performance of the proposed scheme, numerical studies are performed on a set of synthetic test signals, and excellent results have been obtained. This paper also presents an illustrative example where two types of heart sounds are classified. The classification error percentage for the heart sounds using the new classifier is only 6.670% as compared to 56.67% when a general correlation-based classifier is used.

The effect of three time-frequency representations on a novel correlation algorithm is studied. By representing a signal in the time-frequency domain, a redundant representation of the signal is obtained. The algorithm presented relies on such redundancies to extrapolate some significant features of the signal. The developed scheme has been applied to heart sound analysis using real recordings from patients, where the opening snap (OS) is distinguished from the third heart sound (S3). The results for the three time-frequency transforms are compared and very encouraging results have been obtained with S-transform.