top of page

Science Behind

Parky Research

Although there is still way to go to make things perfect, the results show significant success on detecting and differentiating tremor and dyskinesia. 12 patients attended the research and each of them spent around 5,5 hours inside the clinic. There was a task list for patients starting with UPDRS exam and following daily life errands such as pouring water in a glass, preparing a sandwich, washing dishes, buttoning up a shirt etc. During the whole time, 4 cameras recorded the scene. Each 5-hour patient videos had been divided into 30 seconds EPOCHs which counts around 7000 EPOCHs for the whole study. Each of these 30 second videos were analyzed and labeled by two blind Parkinson’s Disease specialist neurologists based on the symptoms occur. The common cluster of two specialists’ opinions (which was around 80%) had been taken into consideration for machine learning techniques. The sensor data and labels by the clinicians had been processed and create the final algorithm. The results were between 70%-80% accuracy based on the limb and symptom that was analyzed. Now Parky study has leveled up to a validation study for the algorithm with wider range of users which is currently going on.

During this research, there was also another ongoing validation study conducted by Apple for the algorithm MM4PD. MM4PD was developed solely for the use with Apple Watch. (1) The study were published on February 3, 2021 on Science Translational Medicine Magazine and the results were shaping the current state of technology.

Parky Research

The design of the study consists 2 different parts, one is in clinic, short term and monitoring in detail and the other one is free-living, long-term and measuring the effects.


Smartwatch sensor data were mapped to MDS-UPDRS ratings by designing an algorithm that worked in bounded conditions such as in-clinic cognitive distraction tasks and further enriched with shorter free-living periods spanning weeks. In the validation phase, motor fluctuations from smartwatch symptom profiles were retrospectively compared to a patient’s prescribed medication times. Longitudinal datasets were divided into design and hold-out sets. Three studies were performed to map sensor data to MDS-UPDRS ratings: (i) the pilot study with 118 patients with Parkinson’s disease (PD) with multiple expert ratings, (ii) the longitudinal patient study with 6+ months of data from 225 patients with PD, and (iii) the longitudinal control study with 171 elderly, non-PD controls.

Parky Research

The clinician reviewed the smartwatch symptom profiles of 112 subjects in the longitudinal patient study who underwent treatment changes. (A) Symptom changes matched the clinician’s expectation of the prescribed medication change in 94% of cases. Unexpected cases revealed plausible incidence of known side effects to medications. (B) Three blinded movement disorder specialists classified 10 sets of profiles as pre- or posttreatment using only the patient’s medication schedule and MDS-UPDRS tremor and dyskinesia ratings from the intake visit; 87.5% of classifications were correct; three misclassifications occurred because raters presumed that an alternate medication had a dominant effect. Six cases were deemed inconclusive and were excluded.


Based on this validation study, the current version of Parky uses the MM4PD algorithm through Apple Watch devices and iPhones. As long as the user wears Apple Watch, the symptoms are tracked 7/24 and can be reported. Tremor and dyskinesia results are classified based on their nature differently. Tremor symptoms has been classified as Unknown, None, Slight, Mild, Moderate, Strong. 

Dyskinesia episodes are rated as likely or unlikely. Parky integrated this top-level algorithm into an end user Parkinson’s Disease patient product by focusing on ease of use, technical accuracy, and user friendliness. Other that reporting the symptoms, Parky also reports detected falls and medication compliance.

As a research platform Parky also has handy tools for both the patient and caregiver to fight against freeze of gaits. Visual cues are known to help for freezing episodes (2) (3) The augmented reality technology has been used to create visual cues that might help trigger the movement for the patients. The auditory cues such as metronome sound are also integrated inside the app which is also debated for many years to apply for Parkinson’s Disease patients who suffers freezing.


Wearable technologies, AI methods has a lot of potential to create a platform where disease management would be much more easier for the case of Parkinson’s. (4) Although there is still way to go, symptom tracking for movement disorders can light up the way in patient and clinician communication and provide a new way of disease management.

Science Behind

Parky technology is based on Parkinson’s Disease clinical research, wearable devices and machine learning methodologies. The main purpose of Parky technology is to provide tools to track the condition for each user 7/24. For Parkinson’s Disease patients, caregivers and healthcare professionals, one of the biggest concerns is to understand the pace and course of the disease. Each example has its own nature and especially for the early diagnosis stages, it is a burden to define the best treatment protocol with minimum trial and error.


Typical symptoms of Parkinson’s disease tell a lot about the course of the condition when well understood and well reported. However within everyday life, it is critically hard for patients to listen to themselves and understand the differences of the movement problems 7/24. In addition to this hurdle, the long periods between doctor visits (especially in times such as pandemic or for immobile patients) cracks the communication between the clinician and the patient. End of the day there often stays a troubled patient and a clinician with very little information trying to pull up the pieces in order to see the overall picture of the condition.


As a movement disorder, Parkinson’s disease symptoms have the potential to be deeply analyzed through motion sensors such as 3d accelerometer which happens to be integrated in almost all types of smartwatches, pedometers etc. The current state of wearable and mobile technologies enables access to vast amount of movement data that needs to be analyzed and interpreted.


In the past few years, there has been huge effort and research conducted to make sense of the movement sensor data and set a standard algorithm for understanding the nature of tremor and dyskinesia episodes. One of those researches has been concluded by Parky team under the roof of H2O Wearhouse. The results are summarized below.

bottom of page