Crowd-sourced machine learning prediction of long COVID using data from the National COVID Cohort Collaborative.

BACKGROUND: While many patients seem to recover from SARS-CoV-2 infections, many patients report experiencing SARS-CoV-2 symptoms for weeks or months after their acute COVID-19 ends, even developing new symptoms weeks after infection. These long-term effects are called post-acute sequelae of SARS-CoV-2 (PASC) or, more commonly, Long COVID. The overall prevalence of Long COVID is currently unknown, and tools are needed to help identify patients at risk for developing long COVID. METHODS: A working group of the Rapid Acceleration of Diagnostics-radical (RADx-rad) program, comprised of individuals from various NIH institutes and centers, in collaboration with REsearching COVID to Enhance Recovery (RECOVER) developed and organized the Long COVID Computational Challenge (L3C), a community challenge aimed at incentivizing the broader scientific community to develop interpretable and accurate methods for identifying patients at risk of developing Long COVID. From August 2022 to December 2022, participants developed Long COVID risk prediction algorithms using the National COVID Cohort Collaborative (N3C) data enclave, a harmonized data repository from over 75 healthcare institutions from across the United States (U.S.). FINDINGS: Over the course of the challenge, 74 teams designed and built 35 Long COVID prediction models using the N3C data enclave. The top 10 teams all scored above a 0.80 Area Under the Receiver Operator Curve (AUROC) with the highest scoring model achieving a mean AUROC of 0.895. Included in the top submission was a visualization dashboard that built timelines for each patient, updating the risk of a patient developing Long COVID in response to clinical events. INTERPRETATION: As a result of L3C, federal reviewers identified multiple machine learning models that can be used to identify patients at risk for developing Long COVID. Many of the teams used approaches in their submissions which can be applied to future clinical prediction questions. FUNDING: Research reported in this RADx® Rad publication was supported by the National Institutes of Health. Timothy Bergquist, Johanna Loomba, and Emily Pfaff were supported by Axle Subcontract: NCATS-STSS-P00438.

Errors elicit frontoparietal theta-gamma coupling that is modulated by endogenous estradiol levels.

Cognitive control-related error monitoring is intimately involved in behavioral adaptation, learning, and individual differences in a variety of psychological traits and disorders. Accumulating evidence suggests that a focus on women's health and ovarian hormones is critical to the study of such cognitive brain functions. Here we sought to identify a novel index of error monitoring using a time-frequency based phase amplitude coupling (t-f PAC) measure and examine its modulation by endogenous levels of estradiol in females. Forty-three healthy, naturally cycling young adult females completed a flanker task while continuous electroencephalogram was recorded on four occasions across the menstrual cycle. Results revealed significant error-related t-f PAC between theta phase generated in fronto-central areas and gamma amplitude generated in parietal-occipital areas. Moreover, this error-related theta-gamma coupling was enhanced by endogenous levels of estradiol both within females across the cycle as well as between females with higher levels of average circulating estradiol. While the role of frontal midline theta in error processing is well documented, this paper extends the extant literature by illustrating that error monitoring involves the coordination between multiple distributed systems with the slow midline theta activity modulating the power of gamma-band oscillatory activity in parietal regions. They further show enhancement of inter-regional coupling by endogenous estradiol levels, consistent with research indicating modulation of cognitive control neural functions by the endocrine system in females. Together, this work identifies a novel neurophysiological marker of cognitive control-related error monitoring in females that has implications for neuroscience and women's health.

Community detection in multi-frequency EEG networks.

Functional connectivity networks of the human brain are commonly studied using tools from complex network theory. Existing methods focus on functional connectivity within a single frequency band. However, it is well-known that higher order brain functions rely on the integration of information across oscillations at different frequencies. Therefore, there is a need to study these cross-frequency interactions. In this paper, we use multilayer networks to model functional connectivity across multiple frequencies, where each layer corresponds to a different frequency band. We then introduce the multilayer modularity metric to develop a multilayer community detection algorithm. The proposed approach is applied to electroencephalogram (EEG) data collected during a study of error monitoring in the human brain. The differences between the community structures within and across different frequency bands for two response types, i.e. error and correct, are studied. The results indicate that following an error response, the brain organizes itself to form communities across frequencies, in particular between theta and gamma bands while a similar cross-frequency community formation is not observed following the correct response.

Assessment of dynamic phase amplitude coupling using matching pursuit.

BACKGROUND: Neuronal transmission and communication are enabled by the interactions across multiple oscillatory frequencies. Phase amplitude coupling (PAC) quantifies these interactions during cognitive brain functions. PAC is defined as the modulation of the amplitude of the high frequency rhythm by the phase of the low frequency rhythm. Existing PAC measures are limited to quantifying the average coupling within a time window of interest. However, as PAC is dynamic, it is necessary to quantify time-varying PAC. Existing time-varying PAC approaches are based on using a sliding window approach. These approaches do not adapt to the signal dynamics, and thus the arbitrary selection of the window length substantially hampers PAC estimation. NEW METHOD: To address the limitations of sliding window PAC estimation approaches, in this paper, we introduce a dynamic PAC measure that relies on matching pursuit (MP). This approach decomposes the signal into time and frequency localized atoms that best describe the signal. Dynamic PAC is quantified by computing the coupling between these time and frequency localized atoms. As such, the proposed approach is data-driven and tracks the change of PAC with time. We evaluate the proposed method on both synthesized and real electroencephalogram (EEG) data. RESULTS: The results from synthesized data show that the proposed method detects the coupled frequencies and the time variation of the coupling correctly with high time and frequency resolution. The analysis of EEG data revealed theta-gamma and alpha-gamma PAC during response and post-response time intervals. COMPARISON WITH EXISTING METHOD(S): Compared to the existing sliding window based approach, the proposed MP based dynamic PAC measure is more effective at capturing PAC within a short time window and is more robust to noise. This is because this method quantifies the low frequency phase and high frequency amplitude components from the time and frequency localized MP atoms and, as such, can capture the signal dynamics. CONCLUSIONS: We posit that the proposed MP based data-driven approach offers a more robust and possibly more sensitive method to effectively quantify and track dynamic PAC.

Graph Theoretic Analysis of Multilayer EEG Connectivity Networks.

Over the past twenty years, functional connectivity of the human brain has been studied in detail using tools from complex network theory. These methods include graph theoretic metrics ranging from the micro-scale such as the degree of a node to the macro-scale such as the small worldness of the brain network. However, most of these network models focus on average activity within a time window of interest and given frequency band. Therefore, they cannot capture the changes in network connectivity across time and different frequency bands. Recently, multilayer brain networks have attracted a lot of attention as they can capture the full view of neuronal connectivity. In this paper, we introduce a multilayer view of the functional connectivity network of the brain, where each layer corresponds to a different frequency band. We construct multi-frequency connectivity networks from electroencephalogram data where the intra-layer edges are quantified by phase synchrony while the inter-layer edges are quantified by phase-amplitude coupling. We then introduce multilayer degree, participation coefficient and clustering coefficient to quantify the centrality of nodes across frequency layers and to identify the importance of different frequency bands. The proposed framework is applied to electroencephalogram data collected during a study of error monitoring in the human brain.

Multivariate Analysis of Bivariate Phase-Amplitude Coupling in EEG Data Using Tensor Robust PCA.

Cross-frequency coupling is emerging as a crucial mechanism that coordinates the integration of spectrally and spatially distributed neuronal oscillations. Recently, phase-amplitude coupling, a form of cross-frequency coupling, where the phase of a slow oscillation modulates the amplitude of a fast oscillation, has gained attention. Existing phase-amplitude coupling measures are mostly confined to either coupling within a region or between pairs of brain regions. Given the availability of multi-channel electroencephalography recordings, a multivariate analysis of phase amplitude coupling is needed to accurately quantify the coupling across multiple frequencies and brain regions. In the present work, we propose a tensor based approach, i.e., higher order robust principal component analysis, to identify response-evoked phase-amplitude coupling across multiple frequency bands and brain regions. Our experiments on both simulated and electroencephalography data demonstrate that the proposed multivariate phase-amplitude coupling method can capture the spatial and spectral dynamics of phase-amplitude coupling more accurately compared to existing methods. Accordingly, we posit that the proposed higher order robust principal component analysis based approach filters out the background phase-amplitude coupling activity and predominantly captures the event-related phase-amplitude coupling dynamics to provide insight into the spatially distributed brain networks across different frequency bands.

Time-Frequency Based Phase-Amplitude Coupling Measure For Neuronal Oscillations.

Oscillatory activity in the brain has been associated with a wide variety of cognitive processes including decision making, feedback processing, and working memory. The high temporal resolution provided by electroencephalography (EEG) enables the study of variation of oscillatory power and coupling across time. Various forms of neural synchrony across frequency bands have been suggested as the mechanism underlying neural binding. Recently, a considerable amount of work has focused on phase-amplitude coupling (PAC)- a form of cross-frequency coupling where the amplitude of a high frequency signal is modulated by the phase of low frequency oscillations. The existing methods for assessing PAC have some limitations including limited frequency resolution and sensitivity to noise, data length and sampling rate due to the inherent dependence on bandpass filtering. In this paper, we propose a new time-frequency based PAC (t-f PAC) measure that can address these issues. The proposed method relies on a complex time-frequency distribution, known as the Reduced Interference Distribution (RID)-Rihaczek distribution, to estimate both the phase and the envelope of low and high frequency oscillations, respectively. As such, it does not rely on bandpass filtering and possesses some of the desirable properties of time-frequency distributions such as high frequency resolution. The proposed technique is first evaluated for simulated data and then applied to an EEG speeded reaction task dataset. The results illustrate that the proposed time-frequency based PAC is more robust to varying signal parameters and provides a more accurate measure of coupling strength.

Graph-to-signal transformation based classification of functional connectivity brain networks.

Complex network theory has been successful at unveiling the topology of the brain and showing alterations to the network structure due to brain disease, cognitive function and behavior. Functional connectivity networks (FCNs) represent different brain regions as the nodes and the connectivity between them as the edges of a graph. Graph theoretic measures provide a way to extract features from these networks enabling subsequent characterization and discrimination of networks across conditions. However, these measures are constrained mostly to binary networks and highly dependent on the network size. In this paper, we propose a novel graph-to-signal transform that overcomes these shortcomings to extract features from functional connectivity networks. The proposed transformation is based on classical multidimensional scaling (CMDS) theory and transforms a graph into signals such that the Euclidean distance between the nodes of the network is preserved. In this paper, we propose to use the resistance distance matrix for transforming weighted functional connectivity networks into signals. Our results illustrate how well-known network structures transform into distinct signals using the proposed graph-to-signal transformation. We then compute well-known signal features on the extracted graph signals to discriminate between FCNs constructed across different experimental conditions. Based on our results, the signals obtained from the graph-to-signal transformation allow for the characterization of functional connectivity networks, and the corresponding features are more discriminative compared to graph theoretic measures.

A Novel EEG Based Spectral Analysis of Persistent Brain Function Alteration in Athletes with Concussion History.

The neurocognitive sequelae of a sport-related concussion and its management are poorly defined. Detecting deficits are vital in making a decision about the treatment plan as it can persist one year or more following a brain injury. The reliability of traditional cognitive assessment tools is debatable, and thus attention has turned to assessments based on electroencephalogram (EEG) to evaluate subtle post-concussive alterations. In this study, we calculated neurocognitive deficits combining EEG analysis with three standard post-concussive assessment tools. Data were collected for all testing modalities from 21 adolescent athletes (seven concussive and fourteen healthy) in three different trials. For EEG assessment, along with linear frequency-based features, we introduced a set of time-frequency (Hjorth Parameters) and nonlinear features (approximate entropy and Hurst exponent) for the first time to explore post-concussive deficits. Besides traditional frequency-band analysis, we also presented a new individual frequency-based approach for EEG assessment. While EEG analysis exhibited significant discrepancies between the groups, none of the cognitive assessment resulted in significant deficits. Therefore, the evidence from the study highlights that our proposed EEG analysis and markers are more efficient at deciphering post-concussion residual neurocognitive deficits and thus has a potential clinical utility of proper concussion assessment and management.

Gait speed estimation using Kalman Filtering on inertial measurement unit data.

Gait speed measurement is vital for diagnosis of motor disorder and monitoring the progress of patient rehabilitation. This study presents an algorithm for moderate distance gait speed measurement from data acquired with inertial motion sensors comprised of a tri-axial accelerometer and a tri-axial gyroscope. Gait speed was measured in four different speed levels set by a treadmill: 0.5, 1, 2, and 3 miles/hour. The calculated speed was tuned by implementing Kalman Filter. The performance of the proposed algorithm was evaluated by calculating the mean square error between estimated speed and the actual treadmill speed. The preliminary results obtained from various treadmill speeds suggest that proposed algorithm estimated speed in a reasonable accuracy. The average error rate was 0.23 m/h which is nearly similar to other studies in this area. Algorithm performance evaluation for various speeds implied that the best performance was exhibited when the speed was set at 1 mile/hour. Moreover, the use of Kalman Filter helped to fine-tune the estimated speed by removing uncertainty and eventually provided a better approximation of the speed measured from the inertial measurement unit.

Evidence of brain functional deficits following sport-related mild traumatic brain injury.

Sport related mild traumatic brain injury (mTBI), generally known as a concussion, is a worldwide critical public health concern nowadays. Despite growing concern emphasized by scientific research and recent media presentation regarding mTBI and its effect in athletics life, the management, and prevention of mTBI are still not properly done. The evaluation mainly hampered due to the lack of proper knowledge, subjective nature of assessment tools including the fact that the brain functional deficits after mTBI can be mild or hidden. As a result, development of an effective tool for proper management of these mild incidents is a subject of active research. In this paper, to examine the neural substrates following mTBI, an analysis based on electroencephalogram (EEG) from twenty control and twenty concussed athletes is presented. Preliminary results suggest that the concussed athletes have a significant increase in delta, theta and alpha power but a decrease in beta power. We also calculated the power for individual frequencies from 1 Hz to 40 Hz in order to find out the specific frequencies with the highest deficits. The significant deficiencies were found at 1-2 Hz of delta band, 6-7 Hz of theta band, 8-10 Hz of the alpha band, and 16-18 Hz and 24-29 Hz of the beta band. Though there was no significant difference as observed in gamma band, we found the deficit was significant at 34-36 Hz range within the gamma band. The observed deficits at various frequencies demonstrate that even if there is no significant difference in the traditional frequency bands, there may be hidden deficits at some specific frequencies within a frequency band. These preliminary results suggest that the EEG analysis at each unity frequency may be more promising means of identifying the neuronal damage than the traditional frequency band based analysis. Eventually, the proposed analysis can provide an improved approximation to monitor the pathophysiological recovery after a concussion.

Automatic diagnosis of melanoma using linear and nonlinear features from digital image.

Melanoma is the most serious type of skin cancer and causes more deaths than other forms of skin cancer. It is a tiny small malignant mole that is usually black or brown but also appears in other color patterns. Early detection of melanoma is key as this is the time period when it is most likely to be cured. Due to the advancement of smartphone technology, automatic and efficient detection of melanoma mole using a smartphone is an active area of research. In this study, we developed an automatic melanoma diagnosis system using images captured from the digital camera. Our work differs from other studies in the area of segmentation of melanoma region and consideration of non-linear features for classification of malignant and benign melanoma. In this paper, a combination of Otsu and k-means clustering segmentation methods are applied to automatically segment and extract the borders of affected region with satisfactory accuracy. Also, we explored and extracted different non-linear features along with color and texture features existed in literature from the lesion mole. The effectiveness of these features was predicted with a machine learning model consisting of five different classifiers. Our model predicted the diagnosis of mole with an accuracy of 89.7%, i.e., around 10% more than reported results by others (to the best of our knowledge) with the same database.

Vertical ground reaction force marker for Parkinson's disease.

Parkinson's disease (PD) patients regularly exhibit abnormal gait patterns. Automated differentiation of abnormal gait from normal gait can serve as a potential tool for early diagnosis as well as monitoring the effect of PD treatment. The aim of current study is to differentiate PD patients from healthy controls, on the basis of features derived from plantar vertical ground reaction force (VGRF) data during walking at normal pace. The current work presents a comprehensive study highlighting the efficacy of different machine learning classifiers towards devising an accurate prediction system. Selection of meaningful feature based on sequential forward feature selection, the swing time, stride time variability, and center of pressure features facilitated successful classification of control and PD gaits. Support Vector Machine (SVM), K-nearest neighbor (KNN), random forest, and decision trees classifiers were used to build the prediction model. We found that SVM with cubic kernel outperformed other classifiers with an accuracy of 93.6%, the sensitivity of 93.1%, and specificity of 94.1%. In comparison to other studies, utilizing same dataset, our designed prediction system improved the classification performance by approximately 10%. The results of the current study underscore the ability of the VGRF data obtained non-invasively from wearable devices, in combination with a SVM classifier trained on meticulously selected features, as a tool for diagnosis of PD and monitoring effectiveness of therapy post pathology.

Preliminary results of residual deficits observed in athletes with concussion history: combined EEG and cognitive study.

Assessment, treatment, and management of sport-related concussions are a widely recognized public health issue. Although several neuropsychological and motor assessment tools have been developed and implemented for sports teams at various levels and ages, the sensitivity of these tests has yet to be validated with more objective measures to make return-to-play (RTP) decisions more confidently. The present study sought to analyze the residual effect of concussions on a sample of adolescent athletes who sustained one or more previous concussions compared to those who had no concussion history. For this purpose, a wide variety of assessment tools containing both neurocognitive and electroencephalogram (EEG) elements were used. All clinical testing and EEG were repeated at 8 months, 10 months, and 12 months post-injury for both healthy and concussed athletes. The concussed athletes performed poorer than healthy athletes on processing speed and impulse control subtest of neurocognitive test on month 8, but no alterations were marked in terms of visual and postural stability. EEG analysis revealed significant differences in brain activities of concussed athletes through all three intervals. These long-term neurocognitive and EEG deficits found from this ongoing sport-related concussion study suggest that the post-concussion physiological deficits may last longer than the observed clinical recovery.

Automatic detection and severity measurement of eczema using image processing.

Chronic skin diseases like eczema may lead to severe health and financial consequences for patients if not detected and controlled early. Early measurement of disease severity, combined with a recommendation for skin protection and use of appropriate medication can prevent the disease from worsening. Current diagnosis can be costly and time-consuming. In this paper, an automatic eczema detection and severity measurement model are presented using modern image processing and computer algorithm. The system can successfully detect regions of eczema and classify the identified region as mild or severe based on image color and texture feature. Then the model automatically measures skin parameters used in the most common assessment tool called "Eczema Area and Severity Index (EASI)," by computing eczema affected area score, eczema intensity score, and body region score of eczema allowing both patients and physicians to accurately assess the affected skin.