Objective: To integrate wearable sensors into brief clinical assessments and use machine learning to infer HD disease status and predict UHDRS motor subscores for specific motor features.

Background: The Unified Huntington’s Disease Rating Scale (UHDRS) is the primary assessment tool for rating motor features in Huntington’s disease (HD) but remains subjective and limited to observed assessments by experienced raters. Digital measures derived from wearable sensors are promising potential tools to objectively measure disease characteristics over time, for both clinical management and as possible biomarkers for disease progression and surrogate endpoints in clinical trials.

Method: Participants with symptomatic HD and healthy controls wore five sensors recording accelerometry and gyroscopic data over the four limbs and the sacrum. Participants completed walking, sitting, and standing tasks during a single office visit. UHDRS motor scores were recorded for participants with symptomatic HD. A two-stage machine learning method was used to classify participants by HD status and to predict selected UHDRS motor subscores.

Results: 14 participants with symptomatic HD and 14 healthy controls completed the study. Machine learning classification predicted participant disease status with a 96.4% accuracy, 100% sensitivity, and 92.9% specificity in leave-one-out cross-validation. Two models (decision tree and Gaussian process regression) predicted mUHDRS subscores with prediction error below 20% of each score range for gait, finger tapping, and dystonia subscores. These models predicted a composite score combining bradykinesia, rigidity, and chorea subscores with error less than 10% of the score range.

Conclusion: In this study, machine learning classifiers trained on brief datasets were able to discriminate between controls and participants with HD and could accurately predict mUHDRS subscores. These results support the utility of biosensors for objective classification and measurement of abnormal movement characteristics in HD. These methods have potential applications in other neurological diseases, such as ataxias.

References: 1. Huntington Study Group. Unified Huntington’s Disease Rating Scale: Reliability and Consistency. Movement Disorders [online serial]. Epub 1996. Accessed at: http://dx.doi.org/10.1002/mds.870110204. 2. Mestre TA, Forjaz MJ, Mahlknecht P, et al. Rating Scales for Motor Symptoms and Signs in Huntington’s Disease: Critique and Recommendations. Mov Disord Clin Pract. 2018;5:111–117. 3. Tabrizi SJ, Leavitt BR, Landwehrmeyer GB, et al. Targeting Huntingtin Expression in Patients with Huntington’s Disease. N Engl J Med. 2019;380:2307–2316. 4. Dinesh K, Snyder CW, Xiong M, et al. A Longitudinal Wearable Sensor Study in Huntington’s Disease. J Huntingtons Dis [online serial]. Epub 2019 Dec 18. Accessed at: http://dx.doi.org/10.3233/JHD-190375. 5. Bennasar M, Hicks YA, Clinch SP, et al. Automated Assessment of Movement Impairment in Huntington’s Disease. IEEE Trans Neural Syst Rehabil Eng. 2018;26:2062–2069. 6. Andrzejewski KL, Dowling AV, Stamler D, et al. Wearable Sensors in Huntington Disease: A Pilot Study. J Huntingtons Dis. 2016;5:199–206. 7. Gordon MF, Grachev ID, Mazeh I, et al. Quantification of Motor Function in Huntington Disease Patients Using Wearable Sensor Devices. Digital Biomarkers. 2019;3:103–115. 8. Lonini L, Dai A, Shawen N, et al. Wearable Sensors for Parkinson’s Disease: Which Data are Worth Collecting for Training Symptom Detection Models [online]. npj Digital Medicine 2018. Accessed at: http://dx.doi.org/10.1038/s41746-018-0071-z. 9. Jeon H, Lee W, Park H, et al. Automatic Classification of Tremor Severity in Parkinson’s Disease Using a Wearable Device. Sensors [online serial]. 2017;17. Accessed at: http://dx.doi.org/10.3390/s17092067. 10. Tibshirani R. Regression Shrinkage and Selection Via the Lasso. J R Stat Soc Series B Stat Methodol. 1996;58:267–288.

« Back to MDS Virtual Congress 2020

MDS Abstracts - https://www.mdsabstracts.org/abstract/automatic-classification-of-abnormal-movement-in-huntingtons-disease-using-wearable-sensors-and-machine-learning/