Predicting Wrist Posture during Occupational Tasks Using Inertial Sensors and Convolutional Neural Networks.

Current methods for ergonomic assessment often use video-analysis to estimate wrist postures during occupational tasks. Wearable sensing and machine learning have the potential to automate this tedious task, and in doing so greatly extend the amount of data available to clinicians and researchers. A method of predicting wrist posture from inertial measurement units placed on the wrist and hand via a deep convolutional neural network has been developed. This study has quantified the accuracy and reliability of the postures predicted by this system relative to the gold standard of optoelectronic motion capture. Ten participants performed 3 different simulated occupational tasks on 2 occasions while wearing inertial measurement units on the hand and wrist. Data from the occupational task recordings were used to train a convolutional neural network classifier to estimate wrist posture in flexion/extension, and radial/ulnar deviation. The model was trained and tested in a leave-one-out cross validation format. Agreement between the proposed system and optoelectronic motion capture was 65% with κ = 0.41 in flexion/extension and 60% with κ = 0.48 in radial/ulnar deviation. The proposed system can predict wrist posture in flexion/extension and radial/ulnar deviation with accuracy and reliability congruent with published values for human estimators. This system can estimate wrist posture during occupational tasks in a small fraction of the time it takes a human to perform the same task. This offers opportunity to expand the capabilities of practitioners by eliminating the tedium of manual postural assessment.

Detecting subject-specific fatigue-related changes in lifting kinematics using a machine learning approach.

Individual responses to fatigue have been observed in lifting kinematics, suggesting a subject-specific approach is necessary for fatigue identification. One-class support vector machines (OCSVM) may provide an objective method to classify fatigue-related kinematic changes during repetitive lifting. Participants completed a repetitive lifting protocol while motion capture recorded lifting motions. Subject-specific kinematics from participants' first 35% of lifts trained OCSVM decision boundaries. The remaining lifts were separated into test sets and classified against the decision boundary to identify the percentage of outlier lifts within each test set. Spearman's correlation assessed if the test sets' percentage of outlier lifts increased concurrently with participants' rating of perceived exertion (RPE). Significant positive associations were found for participants who demonstrated evidence of fatigue, while no significant associations were found for participants who did not demonstrate evidence of fatigue. These results demonstrate the prospective efficacy of an outlier detection tool for fatigue detection during repetitive lifting.Practitioner Summary: An objective subject-specific fatigue detection method is desired for workplace tasks, such as lifting. An outlier detection machine learning approach was identified when lifting movement patterns changed from baseline throughout a repetitive lifting protocol. Participants who demonstrated an increase in outlier movement patterns had a concurrent increase in self-reported fatigue.

Machine Learning in Modeling of Mouse Behavior.

Mouse behavior is a primary outcome in evaluations of therapeutic efficacy. Exhaustive, continuous, multiparametric behavioral phenotyping is a valuable tool for understanding the pathophysiological status of mouse brain diseases. Automated home cage behavior analysis produces highly granulated data both in terms of number of features and sampling frequency. Previously, we demonstrated several ways to reduce feature dimensionality. In this study, we propose novel approaches for analyzing 33-Hz data generated by CleverSys software. We hypothesized that behavioral patterns within short time windows are reflective of physiological state, and that computer modeling of mouse behavioral routines can serve as a predictive tool in classification tasks. To remove bias due to researcher decisions, our data flow is indifferent to the quality, value, and importance of any given feature in isolation. To classify day and night behavior, as an example application, we developed a data preprocessing flow and utilized logistic regression (LG), support vector machines (SVM), random forest (RF), and one-dimensional convolutional neural networks paired with long short-term memory deep neural networks (1DConvBiLSTM). We determined that a 5-min video clip is sufficient to classify mouse behavior with high accuracy. LG, SVM, and RF performed similarly, predicting mouse behavior with 85% accuracy, and combining the three algorithms in an ensemble procedure increased accuracy to 90%. The best performance was achieved by combining the 1DConv and BiLSTM algorithms yielding 96% accuracy. Our findings demonstrate that computer modeling of the home-cage ethome can clearly define mouse physiological state. Furthermore, we showed that continuous behavioral data can be analyzed using approaches similar to natural language processing. These data provide proof of concept for future research in diagnostics of complex pathophysiological changes that are accompanied by changes in behavioral profile.

Detection of COVID-19 from chest x-ray images using transfer learning.

Purpose: The objective of this study is to develop and evaluate a fully automated, deep learning-based method for detection of COVID-19 infection from chest x-ray images. Approach: The proposed model was developed by replacing the final classifier layer in DenseNet201 with a new network consisting of global averaging layer, batch normalization layer, a dense layer with ReLU activation, and a final classification layer. Then, we performed an end-to-end training using the initial pretrained weights on all the layers. Our model was trained using a total of 8644 images with 4000 images each in normal and pneumonia cases and 644 in COVID-19 cases representing a large real dataset. The proposed method was evaluated based on accuracy, sensitivity, specificity, ROC curve, and F 1 -score using a test dataset comprising 1729 images (129 COVID-19, 800 normal, and 800 pneumonia). As a benchmark, we also compared the results of our method with those of seven state-of-the-art pretrained models and with a lightweight CNN architecture designed from scratch. Results: The proposed model based on DenseNet201 was able to achieve an accuracy of 94% in detecting COVID-19 and an overall accuracy of 92.19%. The model was able to achieve an AUC of 0.99 for COVID-19, 0.97 for normal, and 0.97 for pneumonia. The model was able to outperform alternative models in terms of overall accuracy, sensitivity, and specificity. Conclusions: Our proposed automated diagnostic model yielded an accuracy of 94% in the initial screening of COVID-19 patients and an overall accuracy of 92.19% using chest x-ray images.

Improved MUP Template Estimation Using Local Time Warping and Kernel Weighted Averaging.

A motor unit potential (MUP) template, which represents the shapes of the MUPs within a MUP train, provides information related to the morphology and physiology of the sampled motor unit. This work presents an improved MUP template estimation technique that uses local time warping and kernel weighted ensemble averaging. An analysis of the algorithm, and a description of the improvements compared with spike triggered averaging is given. MUP template estimates were evaluated using simulated EMG signals with a known gold standard template for each motor unit potential train. Statistically significant reduction in template estimation error is shown, both within the baseline and duration portions of a MUP.

Correction: Effects of Aging on Genioglossus Motor Units in Humans.

[This corrects the article DOI: 10.1371/journal.pone.0104572.].

Motor unit number estimates and neuromuscular transmission in the tibialis anterior of master athletes: evidence that athletic older people are not spared from age-related motor unit remodeling.

Muscle motor unit numbers decrease markedly in old age, while remaining motor units are enlarged and can have reduced neuromuscular junction transmission stability. However, it is possible that regular intense physical activity throughout life can attenuate this remodeling. The aim of this study was to compare the number, size, and neuromuscular junction transmission stability of tibialis anterior (TA) motor units in healthy young and older men with those of exceptionally active master runners. The distribution of motor unit potential (MUP) size was determined from intramuscular electromyographic signals recorded in healthy male Young (mean ± SD, 26 ± 5 years), Old (71 ± 4 years) and Master Athletes (69 ± 3 years). Relative differences between groups in numbers of motor units was assessed using two methods, one comparing MUP size and muscle cross-sectional area (CSA) determined with MRI, the other comparing surface recorded MUPs with maximal compound muscle action potentials and commonly known as a "motor unit number estimate (MUNE)". Near fiber (NF) jiggle was measured to assess neuromuscular junction transmission stability. TA CSA did not differ between groups. MUNE values for the Old and Master Athletes were 45% and 40%, respectively, of the Young. Intramuscular MUPs of Old and Master Athletes were 43% and 56% larger than Young. NF jiggle was slightly higher in the Master Athletes, with no difference between Young and Old. These results show substantial and similar motor unit loss and remodeling in Master Athletes and Old individuals compared with Young, which suggests that lifelong training does not attenuate the age-related loss of motor units.

Aggregate features in multisample classification problems.

This paper evaluates the classification of multisample problems, such as electromyographic (EMG) data, by making aggregate features available to a per-sample classifier. It is found that the accuracy of this approach is superior to that of traditional methods such as majority vote for this problem. The classification improvements of this method, in conjunction with a confidence measure expressing the per-sample probability of classification failure (i.e., a hazard function) is described and measured. Results are expected to be of interest in clinical decision support system development.

Effects of aging on genioglossus motor units in humans.

The genioglossus is a major upper airway dilator muscle thought to be important in obstructive sleep apnea pathogenesis. Aging is a risk factor for obstructive sleep apnea although the mechanisms are unclear and the effects of aging on motor unit remodeled in the genioglossus remains unknown. To assess possible changes associated with aging we compared quantitative parameters related to motor unit potential morphology derived from EMG signals in a sample of older (n = 11; >55 years) versus younger (n = 29; <55 years) adults. All data were recorded during quiet breathing with the subjects awake. Diagnostic sleep studies (Apnea Hypopnea Index) confirmed the presence or absence of obstructive sleep apnea. Genioglossus EMG signals were analyzed offline by automated software (DQEMG), which estimated a MUP template from each extracted motor unit potential train (MUPT) for both the selective concentric needle and concentric needle macro (CNMACRO) recorded EMG signals. 2074 MUPTs from 40 subjects (mean±95% CI; older AHI 19.6±9.9 events/hr versus younger AHI 30.1±6.1 events/hr) were extracted. MUPs detected in older adults were 32% longer in duration (14.7±0.5 ms versus 11.1±0.2 ms; P  =  0.05), with similar amplitudes (395.2±25.1 µV versus 394.6±13.7 µV). Amplitudes of CNMACRO MUPs detected in older adults were larger by 22% (62.7±6.5 µV versus 51.3±3.0 µV; P<0.05), with areas 24% larger (160.6±18.6 µV.ms versus 130.0±7.4 µV.ms; P<0.05) than those detected in younger adults. These results confirm that remodeled motor units are present in the genioglossus muscle of individuals above 55 years, which may have implications for OSA pathogenesis and aging related upper airway collapsibility.

Characterizing EMG data using machine-learning tools.

Effective electromyographic (EMG) signal characterization is critical in the diagnosis of neuromuscular disorders. Machine-learning based pattern classification algorithms are commonly used to produce such characterizations. Several classifiers have been investigated to develop accurate and computationally efficient strategies for EMG signal characterization. This paper provides a critical review of some of the classification methodologies used in EMG characterization, and presents the state-of-the-art accomplishments in this field, emphasizing neuromuscular pathology. The techniques studied are grouped by their methodology, and a summary of the salient findings associated with each method is presented.

Neurogenic changes in the upper airway of patients with obstructive sleep apnea.

RATIONALE: Controversy persists regarding the presence and importance of hypoglossal nerve dysfunction in obstructive sleep apnea (OSA). OBJECTIVES: We assessed quantitative parameters related to motor unit potential (MUP) morphology derived from electromyographic (EMG) signals in patients with OSA versus control subjects and hypothesized that signs of neurogenic remodeling would be present in the patients with OSA. METHODS: Participants underwent diagnostic sleep studies to obtain apnea-hypopnea indices. Muscle activity was detected with 50-mm concentric needle electrodes. The concentric needle was positioned at more than 10 independent sites per subject, after the local anatomy of the upper airway musculature was examined by ultrasonography. All activity was quantified with subjects awake, during supine eupneic breathing while wearing a nasal mask connected to a pneumotachograph. Genioglossus EMG signals were analyzed offline by automated software (DQEMG), which extracted motor unit potential trains (MUPTs) contributed by individual motor units from the composite EMG signals. Quantitative measurements of MUP templates, including duration, peak-to-peak amplitude, area, area-to-amplitude ratio, and size index, were compared between the untreated patients with OSA and healthy control subjects. MEASUREMENTS AND MAIN RESULTS: A total of 1,655 MUPTs from patients with OSA (n = 17; AHI, 55 ± 6/h) and control subjects (n = 14; AHI, 4 ± 1/h) were extracted from the genioglossus muscle EMG signals. MUP peak-to-peak amplitudes in the patients with OSA were not different compared with the control subjects (397.5 ± 9.0 vs. 382.5 ± 10.0 µV). However, the MUPs of the patients with OSA were longer in duration (11.5 ± 0.1 vs. 10.3 ± 0.1 ms; P < 0.001) and had a larger size index (4.09 ± 0.02 vs. 3.92 ± 0.02; P < 0.001) compared with control subjects. CONCLUSIONS: These results confirm and quantify the extent and existence of structural neural remodeling in OSA.

Validating motor unit firing patterns extracted by EMG signal decomposition.

Motor unit (MU) firing pattern information can be used clinically or for physiological investigation. It can also be used to enhance and validate electromyographic (EMG) signal decomposition. However, in all instances the validity of the extracted MU firing patterns must first be determined. Two supervised classifiers that can be used to validate extracted MU firing patterns are proposed. The first classifier, the single/merged classifier (SMC), determines whether a motor unit potential train (MUPT) represents the firings of a single MU or the merged activity of more than one MU. The second classifier, the single/contaminated classifier (SCC), determines whether the estimated number of false-classification errors in a MUPT is acceptable or not. Each classifier was trained using simulated data and tested using simulated and real data. The accuracy of the SMC in categorizing a train correctly is 99% and 96% for simulated and real data, respectively. The accuracy of the SCC is 84% and 81% for simulated and real data, respectively. The composition of these classifiers, their objectives, how they were trained, and the evaluation of their performances using both simulated and real data are presented in detail.

Intramuscular EMG signal decomposition.

Information regarding motor unit potentials (MUPs) and motor unit fi ring patterns during muscle contractions is useful for physiological investigation and clinical examinations either for the understanding of motor control or for the diagnosis of neuromuscular disorders. In order to obtain such information, composite electromyographic (EMG) signals are decomposed (i.e., resolved into their constituent motor unit potential trains [MUPTs]). The goals of automatic decomposition techniques are to create a MUPT for each motor unit that contributed significant MUPs to the original composite signal. Diagnosis can then be facilitated by decomposing a needle-detected EMG signal, extracting features of MUPTs, and finally analyzing the extracted features (i.e., quantitative electromyography). Herein, the concepts of EMG signals and EMG signal decomposition techniques are explained. The steps involved with the decomposition of an EMG signal and the methods developed for each step, along with their strengths and limitations, are discussed and compared. Finally, methods developed to evaluate decomposition algorithms and assess the validity of the obtained MUPTs are reviewed and evaluated.

A review of clinical quantitative electromyography.

Information regarding the morphology of motor unit potentials (MUPs) and motor unit firing patterns can be used to help diagnose, treat, and manage neuromuscular disorders. In a conventional electromyographic (EMG) examination, a clinician manually assesses the characteristics of needle-detected EMG signals across a number of distinct needle positions and forms an overall impression of the condition of the muscle. Such a subjective assessment is highly dependent on the skills and level of experience of the clinician, and is prone to a high error rate and operator bias. Quantitative methods have been developed to characterize MUP waveforms using statistical and probabilistic techniques that allow for greater objectivity and reproducibility in supporting the diagnostic process. In this review, quantitative EMG (QEMG) techniques ranging from simple reporting of numeric MUP values to interpreted muscle characterizations are presented and reviewed in terms of their clinical potential to improve status quo methods. QEMG techniques are also evaluated in terms of their suitability for use in a clinical decision support system based on previously established criteria. Aspects of prototype clinical decision support systems are then presented to illustrate some of the concepts of QEMG-based decision making.

Evaluation of motor unit placement algorithms for EMG simulation.

Motor unit layout algorithms have a significant effect on motor unit fibre densities recorded. Motor unit fibre densities are affected by both the method used to place the motor unit territories, and the mechanism by which muscle fibres are assigned to motor units. The first of these should emulate the process by which separate motor neurons create overlapping territories that cover the muscle cross section, while the second should have some relation to the processes involved with axonal arborization and development of the spatial dispersion of the neuro-muscular junctions. The success of an algorithm in creating physiologically realistic motor unit layouts may be evaluated, in part, by examining the distribution of the muscle fibres assigned to the motor units. This paper examines the motor unit fibre densities found in muscles created by two recent algorithms, and explores the degree to which the concepts used by these algorithms may be shared.

Assessing motor deficits in compressive neuropathy using quantitative electromyography.

BACKGROUND: Studying the changes that occur in motor unit potential trains (MUPTs) may provide insight into the extent of motor unit loss and neural re-organization resulting from nerve compression injury. The purpose of this study was to determine the feasibility of using decomposition-based quantitative electromyography (DQEMG) to study the pathophysiological changes associated with compression neuropathy. METHODS: The model used to examine compression neuropathy was carpal tunnel syndrome (CTS) due to its high prevalence and ease of diagnosis. Surface and concentric needle electromyography data were acquired simultaneously from the abductor pollicis brevis muscle in six individuals with severe CTS, eight individuals with mild CTS and nine healthy control subjects. DQEMG was used to detect intramuscular MUPTs during constant-intensity contractions and to estimate parameters associated with the surface- and needle-detected motor unit potentials (SMUPs and MUPs, respectively). MUP morphology and stability, SMUP morphology and motor unit number estimates (MUNEs) were compared among the groups using Kruskal-Wallis tests. RESULTS: The severe CTS group had larger amplitude and longer duration MUPs and smaller MUNEs than the mild CTS and control groups, suggesting that the individuals with severe CTS had motor unit loss with subsequent collateral reinnervation, and that DQEMG using a constant-intensity protocol was sensitive to these changes. SMUP morphology and MUP complexity and stability did not significantly differ among the groups. CONCLUSIONS: These results provide evidence that MUP amplitude parameters and MUNEs obtained using DQEMG, may be a valuable tool to investigate pathophysiological changes in muscles affected by compressive motor neuropathy to augment information obtained from nerve conduction studies. Although there were trends in many of these measures, in this study, MUP complexity and stability and SMUP parameters were, of limited value.

Bayesian aggregation versus majority vote in the characterization of non-specific arm pain based on quantitative needle electromyography.

BACKGROUND: Methods for the calculation and application of quantitative electromyographic (EMG) statistics for the characterization of EMG data detected from forearm muscles of individuals with and without pain associated with repetitive strain injury are presented. METHODS: A classification procedure using a multi-stage application of Bayesian inference is presented that characterizes a set of motor unit potentials acquired using needle electromyography. The utility of this technique in characterizing EMG data obtained from both normal individuals and those presenting with symptoms of "non-specific arm pain" is explored and validated. The efficacy of the Bayesian technique is compared with simple voting methods. RESULTS: The aggregate Bayesian classifier presented is found to perform with accuracy equivalent to that of majority voting on the test data, with an overall accuracy greater than 0.85. Theoretical foundations of the technique are discussed, and are related to the observations found. CONCLUSIONS: Aggregation of motor unit potential conditional probability distributions estimated using quantitative electromyographic analysis, may be successfully used to perform electrodiagnostic characterization of "non-specific arm pain." It is expected that these techniques will also be able to be applied to other types of electrodiagnostic data.

Validation of motor unit potential trains using motor unit firing pattern information.

A robust and fast method to assess the validity of a motor unit potential train (MUPT) obtained by decomposing a needle-detected EMG signal is proposed. This method determines whether a MUPT represents the firings of a single motor unit (MU) or the merged activity of more than one MU, and if is a single train it identifies whether the estimated levels of missed and false classification errors in the MUPT are acceptable. Two supervised classifiers, the Single/Merged classifier (SMC) and the Error Rate classifier (ERC), and a linear model for estimating the level of missed classification error have been developed for this objective. Experimental results using simulated data show that the accuracy of the SMC and the ERC in correctly categorizing a train is 99% and %84 respectively.

Physiologically based simulation of clinical EMG signals.

An algorithm that generates electromyographic (EMG) signals consistent with those acquired in a clinical setting is described. Signals are generated using a model constructed to closely resemble the physiology and morphology of skeletal muscle, combined with line source models of commonly used needle electrodes positioned in a way consistent with clinical studies. The validity of the simulation routines is demonstrated by comparing values of statistics calculated from simulated signals with those from clinical EMG studies of normal subjects. The simulated EMG signals may be used to explore the relationships between muscle structure and activation and clinically acquired EMG signals. The effects of motor unit (MU) morphology, activation, and neuromuscular junction activity on acquired signals can be analyzed at the fiber, MU and muscle level. Relationships between quantitative features of EMG signals and muscle structure and activation are discussed.