A wearable sensor patch, worn on the skin, tracks muscle vibrations and movement to help doctors fine-tune Parkinson’s treatments in real time. Parkinson’s disease has doubled in prevalence over the last 25 years, yet treatment decisions still rely on brief, subjective clinician assessments that miss symptoms fluctuating over minutes or hours. Over- or under-prescribing medication or deep brain stimulation settings leaves patients with uncontrolled motor symptoms or severe side effects such as involuntary jerking movements. The problem is that existing wearable sensors, based on inertial measurement units, cannot measure rigidity or reliably link muscle activity to symptom severity. This project will combine inertial sensors with mechanomyography—which detects low-frequency muscle vibrations—into a single wearable device already proven to track all key motor symptoms. Over 27 months, researchers will develop artificial intelligence algorithms that use the sensor data to recommend personalised medication doses for 40 patients and deep brain stimulation settings for 20 patients. If successful, the system could replace subjective trial-and-error titration with objective, data-driven optimisation, reducing disability and improving quality of life for people with Parkinson’s.
View original technical description
Background Prevalence of Parkinson s disease has doubled within the last 25 years, and it is now second most common neurodegenerative disorder and fastest-growing. Patients already suffer from limited time with overburdened medical services; automated and telemedicine approaches are critically needed. Parkinson s treatment, through pharmaceuticals or deep brain stimulation (DBS), must be tailored to specific patients as there is a wide variation in symptom response. This demands precise assessment of the key motor symptoms - bradykinesia, rigidity, and tremor – and subsequent titration of medication or stimulation parameters for optimal response. Clinician assessment is the basis of titration yet allows only a cross-sectional assessment of symptoms which vary rapidly over minutes/hours. It is further liable to significant inter- and intra-rater subjectivity. Over or under prescription renders treatment ineffective (i.e. causing OFF-time, where motor symptoms return) and can amplify harmful side-effects (e.g., Levodopa-induced dyskinesia, that is, uncontrolled hyperactive movements of the head, trunk, and limbs), which greatly affects a patient's quality-of-life. Achieving the best outcomes for people with Parkinson s relies on the skill and availability of the treating clinician(s), leading to a variability in care delivered across the country. Precise symptom quantification coupled with AI powered treatment optimisation suggestions could improve symptom management and reduce disability. Automated Parkinson's assessment has attracted significant research/commercial interest, however, reliance on inertial measurement units (IMUs) limits diagnostic range. Challenges in measurement of rigidity, correlating muscle activation to symptomatic severity, and sensor fusion remain unresolved. Furthermore, automated intelligence successfully mapping symptomatic assessment to titration of medication is not established. We address these issues through an integrated AI-wearable sensor package. Our system integrates IMUs and mechanomyography (MMG), resulting in the first wearable proven to track all the key motor symptoms of Parkinson s (IEEE Trans Neural/Rehab Eng, 2020). The patented wearable enables robust use out-of-clinic in an inexpensive, reusable package (IEEE Trans Mechatronics, 2020;Finalist: Best Paper). Research Objectives Published results and awarded IP have proven diagnostic feasibility. However, the unique data produced by our wearable provides an opportunity to create machine-learning algorithms for patient-specific titration of medication, representing a revolutionary step in Parkinson's management. This 27-month programme of works will develop an AI system with this capacity. Work will: Develop AI algorithms optimising deep brain stimulation (DBS)-parameters on Parkinson's patients in-clinic (n=20); Extend the AI to titrate pharmaceutical medication in clinic (n=40); Deliver a sub-study, focussed on understanding whether capturing data on a patient's nonmotor symptoms, via a smart wearable, may enhance the value/accuracy of our platform. Methods The in-clinic experimental session includes clinician s-based evaluation and device monitoring before and up to 4-hours after a levodopa challenge test (administration of a levodopa dose) with the addition of ON-and-OFF stimulation states in the DBS-group. Extensive patient input will be sought at all stages of project delivery. We will form patient focus groups that will meet throughout the programme to inform design, judge control gates, and evaluate prototypes. AI development will be intensive throughout the project with multiple rounds of iteration and prototyping anticipated, requiring significant clinical translation across the consortium.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know