Objective: To quantify the PD motor syndrome with increased temporal and severity granularity using non-invasive internet-of-things (IoT) devices and deep learning methods.

Background: Brady-hypokinetic parkinsonistic and hyperkinetic dyskinetic spontaneous movements are the fundamental characteristic of the motor syndrome of patients with Parkinson’s disease (PwP). These motor phenomena are usually assessed by clinical rating scales on the basis of few discernible severity levels mostly only few times per day. New technologies could be leveraged to increase the temporal and severity resolution for diagnostic and therapeutic purposes.

Methods: Ethical approval was obtained prior to experimentation. We evaluated 30 PwP with progressed PD and recorded about 8800 minutes of labeled raw movement data in a free-living setting acquired from a wrist-worn 6-axis accelerometer/gyroscope. Clinical information was assessed every minute by a certified movement disorder expert. Sensor data was cleaned, preprocessed, and windows of five seconds were used to train a deep convolutional neural network (CNN) to predict the motor behavior on an ordered 9-step discrete PD9 scale. The PD9 ranges from severe bradyhypokinesia [-4] to severe dyskinesia [+4]. Statistical smoothing procedures were applied post processing.

Results: The PD9 quantifies both the loss and the excess of spontaneous movement as seen in the motor syndrome of PwP. It can be recorded with high granularity for temporal resolution. The scale is based on the principles of standard clinical evaluation, and meets the needs and capabilities of current data science and IoT technology. The validity of the novel scale to granularly measure the PD motor syndrome was cross-validated using standardized clinical items showing excellent correlation on subject level with R of 0.91.

Conclusions: We describe the objective quantification of PwP including both hypokinetic and hyperkinetic phenomena using deep learning methodology applied to objective inertial measurement data. Our method is not limited to a controlled test setup, but can be applied in free-living situations. The high temporal resolution in the range of seconds to minutes could be used for numerous clinical indications, such as monitoring of patients, diagnosis, and closed-loop therapy.

« Back to 2018 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/quantification-of-the-pd-motor-syndrome-using-an-inclusive-score-pd9-derived-from-sensor-data-with-deep-learning/