248 Tremor frequency can be measured using smartphone video

Textbook descriptions of tremulous disorders typically mention the tremor frequencies characteristic of particular conditions, and determination of frequency can aid diagnosis. However, current clinical practice does not involve routine accelerometry or EMG for most tremulous patients, or use of smartphone accel- erometer, and the human eye cannot accurately measure tremor frequency. 37 hand videos were recorded with smartphone camera from 5 essential tremor participants, 10 Parkin- son’s participants and 1 functional tremor participant, either showing a hand resting or held in posture. Tremor was simiultaneously measured using a commercially available accelerometer for clinical use (‘Natus Neurology’). The computing technique of ‘optical flow’ was applied to the smartphone video to measure pixel movement over time in two directions perpendicular to the long axis of the hand. Dominant frequencies were extracted from smartphone video optical flow and from clinic accelerometer data using fast Fourier transform. Bland-Altman analysis showed a mean of the difference in measurements between smartphone video (pixel optical flow) and accelerometer of 0.05 Hz (SD 0.16 Hz) with 95% confidence intervals for this mean difference of -0.26 Hz to +0.36 Hz. We demonstrate excellent agreement between our computing method for smartphone video tremor measurement and the gold standard of accelerometry. Our computing technique suggests that neu- rologists already have ‘point and press’ contactless equipment in their pocket that has the potential to augment routine clinical assessment by measuring tremor frequency. stefanwilliams@doctors.org.uk