In the course of Parkinson's disease, nerve cells are degraded in the brain. This primarily affects the movement centre. Among other things, this leads to a slowing down of movement sequences, leg blockages when crossing thresholds and trembling of the hands, legs and head. In addition, the disease causes numerous other complaints such as pain and mood swings, memory or sleep disorders. For those affected - over 15,000 people in Switzerland alone - medication can significantly improve everyday functions and thus quality of life.
The correct dosage is important here, as medication is associated with side effects. In the advanced stage, the range in which someone responds well to the medication becomes smaller and smaller. Therefore, at this stage, it is crucial that the medication can be dosed accurately. To determine the right amount, the doctor needs a clear picture of the motor symptoms. Researchers at the iHomeLab and the ARTORG Center for Biomedical Engineering Research at the University of Bern have now demonstrated a way in which a so-called cyber-human system can collect and evaluate precisely this information with the help of sensors. This data can be an important decision-making aid for the doctor.
Uncomplicated but precise measurements
Patients today are usually asked to keep a Parkinson's diary, mainly to record tremors by hand. However, especially for people with cognitive impairment or depression - both of which can be effects of the disease - this daily preoccupation with their disease is a burden - and observation is always subjective. The alternative is weekly tests in the laboratory, which are costly and only provide snapshots. The sensor system with the modified smart watch that we developed at the iHomeLab, on the other hand, can continuously and objectively measure how strong the tremor is at any given time. It can not only provide information about the current situation, but possibly also indicate changes in the course of the disease before they become obvious to the patient.
Sophisticated sensor technology
The sensor system is not only extremely precise in its measurement, but also sophisticated: For the patient, two or three sensors with Bluetooth signal on the body - on the wrist, on the belt and on the leg - are sufficient. The strength of the Bluetooth signal to the worn sensor indicates the distance of the patient to the stationary sensors in the room. In the home, sensors are placed in various locations where similar movements are expected on a daily basis, such as by the sink in the kitchen or at doorsteps that can pose a difficulty for people with Parkinson's disease. These sensors are not simple motion detectors - otherwise every additional person or even a cat in the flat would confuse the measurement. The acceleration data on the three sensors worn are then evaluated. The "ADLs", the "activities of daily living", can be measured and compared over time via the localisation with the fixed sensors.
Evaluation with machine learning
The challenge was to intelligently connect a wide variety of sensor data and to develop innovative analysis methods using machine learning. Collecting the data alone is not enough. They are then analysed with the help of machine learning algorithms in order to quantify the Parkinson's symptoms in the course of the day. The measured data are uploaded anonymously to a secure project server for evaluation. Only the attending physician sees the evaluations.