Crowdsourcing digital health measures to predict Parkinson's disease severity: the Parkinson's Disease Digital Biomarker DREAM Challenge.

Consumer wearables and sensors are a rich source of data about patients' daily disease and symptom burden, particularly in the case of movement disorders like Parkinson's disease (PD). However, interpreting these complex data into so-called digital biomarkers requires complicated analytical approaches, and validating these biomarkers requires sufficient data and unbiased evaluation methods. Here we describe the use of crowdsourcing to specifically evaluate and benchmark features derived from accelerometer and gyroscope data in two different datasets to predict the presence of PD and severity of three PD symptoms: tremor, dyskinesia, and bradykinesia. Forty teams from around the world submitted features, and achieved drastically improved predictive performance for PD status (best AUROC = 0.87), as well as tremor- (best AUPR = 0.75), dyskinesia- (best AUPR = 0.48) and bradykinesia-severity (best AUPR = 0.95).

Assessing aberrant muscle activity patterns via the analysis of surface EMG data collected during a functional evaluation.

BACKGROUND: Surface electromyographic (EMG) recordings collected during the performance of functional evaluations allow clinicians to assess aberrant patterns of muscle activity associated with musculoskeletal disorders. This assessment is typically achieved via visual inspection of the surface EMG data. This approach is time-consuming and leads to accurate results only when the assessment is carried out by an EMG expert. METHODS: A set of algorithms was developed to automatically evaluate aberrant patterns of muscle activity. EMG recordings collected during the performance of functional evaluations in 62 subjects (22 to 61 years old) were used to develop and characterize the algorithms. Clinical scores were generated via visual inspection by an EMG expert using an ordinal scale capturing the severity of aberrant patterns of muscle activity. The algorithms were used in a case study (i.e. the evaluation of a subject with persistent back pain following instrumented lumbar fusion who underwent lumbar hardware removal) to assess the clinical suitability of the proposed technique. RESULTS: The EMG-based algorithms produced accurate estimates of the clinical scores. Results were primarily obtained using a linear regression approach. However, when the results were not satisfactory, a regression implementation of a Random Forest was utilized, and the results compared with those obtained using a linear regression approach. The root-mean-square error of the clinical score estimates produced by the algorithms was a small fraction of the ordinal scale used to rate the severity of the aberrant patterns of muscle activity. Regression coefficients and associated 95% confidence intervals showed that the EMG-based estimates fit well the clinical scores generated by the EMG expert. When applied to the clinical case study, the algorithms appeared to capture the characteristics of the muscle activity patterns associated with persistent back pain following instrumented lumbar fusion. CONCLUSIONS: The proposed approach relies on EMG-based measures to generate accurate estimates of the severity of aberrant patterns of muscle activity. The results obtained in the case study suggest that the proposed technique is suitable to derive clinically-relevant information from EMG data collected during functional evaluations.

A novel method for assessing the severity of levodopa-induced dyskinesia using wearable sensors.

Patients with Parkinson's disease often experience significant changes in the severity of dyskinesia when they undergo titration of their medications. Dyskinesia is marked by involuntary jerking movements that occur randomly in a burst-like fashion. The burst-like nature of such movements makes it difficult to estimate the clinical scores of severity of dyskinesia using wearable sensors. Clinical observations are generally made over intervals of 15-30 s. On the other hand, techniques designed to estimate the severity of dyskinesia based on the analysis of wearable sensor data typically use data segments of approximately 5 s. Consequently, some data segments might include dyskinetic movements, whereas others might not. Herein, we propose a novel method suitable to automatically select data segments from the training dataset that are marked by dyskinetic movements. The proposed method also aggregates results derived from the testing dataset using a machine learning algorithm to estimate the severity of dyskinesia from wearable sensor data. Results obtained from the analysis of sensor data collected from seven subjects with Parkinson's disease showed a marked improvement in the accuracy of the estimation of clinical scores of dyskinesia.