Skip to navigation Skip to main content Skip to footer

Approved Research

Automated actigraphy analysis for identification of prodromal Parkinson's disease or related disorders

Principal Investigator: Dr Emmanuel During
Approved Research ID: 97043
Approval date: March 22nd 2023

Lay summary

In many cases, Parkinson's disease (PD) and related disorders are preceded by early sleep disturbances. A characteristic sleep disorder is REM sleep behavior disorder (RBD), which is an abnormal motor disinhibition during REM sleep, causing twitches and dream-enactment episodes. In at least 80 % of RBD cases, PD or a related disorder will manifest within 10 years. The prevalence of RBD is between 1-2% in the general population. An effective screening method of RBD would have high impact in terms of identifying patients at PD risk, when progression of their neurodegenerative disease can be slowed down or halted.

Current screening methods use questionnaires; however, despite their good accuracy, close to 90%, RBD questionnaires do not have sufficient precision for general population screening. Moreover, the gold standard for a definite diagnosis is costly and requires an in-lab sleep test. Our research suggests that actigraphy could be used on its own or as a supplement to questionnaires for screening purposes. Using machine learning, we found that actigraphy could classify RBD correctly in 92.9 % of recordings. Moreover, this model supplies RBD classification scores that indicate disease severity.

In this project, associations between actigraphy-based RBD classification scores, genetics, and various clinical variables of interest. Carrying out this study in the UK Biobank has many advantages to previous studies of RBD, as available datasets based on definite diagnosis are very scarce. Applying our detection algorithm in the UKBB may allow to find new genotypes and phenotypes (described in detail in A4) related to RBD and disease progression.

During this 3-year project, we will: 1) Fine-tune the machine learning-based classifier for data in the UK; 2) Run the RBD classifier on all accelerometer data from suitable participants; 3) Replicate genetic profiles known for PD and definite RBD, in individuals predicted to have RBD using our method; 4) Describe cognitive, motor, autonomic, and biological differences in individuals predicted to have RBD using our method, against individuals predicted to not have RBD

The public health impact of this research would be: 1) to increase our understanding of PD subtypes and disease mechanisms, 2) accelerate discovery of therapies for PD by enhancing recruitment pipelines with the inclusion of participants with prodromal PD, ie with RBD, and 3) offer a screening tool for RBD once neuroprotective therapies for PD become available.