Short communicationRapid tremor frequency assessment with the iPhone accelerometer☆
Introduction
Tremor, a ’rhythmical, involuntary oscillatory movement of a body part’ [1] is the most common movement disorder. Tremor can appear alone as in essential tremor (ET) or as part of other conditions, such as Parkinson’s disease, multiple sclerosis, or cerebellar damage.
Frequency as a property of tremor has been studied for quite some time; Holmes in 1904 identified that tremor frequencies were characteristic to certain conditions, and thought that this might be linked to their pathophysiology [2]. Tremor can be found in a wide variety of frequencies and amplitudes. For example, the frequency of classical rest tremor in Parkinson’s disease is between 4 and 7 Hz with large variations in tremor amplitude [1]. Essential tremor and dystonic tremor have a frequency between 4 and 12 Hz, multiple sclerosis between 2 and 10 Hz, and orthostatic tremor between 13 and 18 Hz [3].
Although clinical rating scales are quite useful for a rough assessment of tremor amplitude, assessing frequency by observation is by definition approximate and inaccurate. Clinical assessments then, are often combined with EMG measurements and accelerometry to determine tremor amplitude and frequency [4].
The ways of measuring tremor range dramatically. Innovative methods that have been used to measure tremor include an electromagnetic tracking device [5], a mechanical linkage device on the fingertip [6], lasers [7], and digitizing tablets [8], to name a few.
Although these various techniques are interesting and useful in their own right, they have demonstrated limited usability in a clinical setting, where flexibility, efficiency, and accuracy are paramount. The use of portable triaxial accelerometers is perhaps the easiest way to assess tremor [9], however this also involves purchase and transportation of necessary equipment from patient to patient. All in all, the existing methods for tremor assessment are quite cumbersome and expensive.
The challenge is finding a universally accessible, easily used device that can give a quick, accurate representation of tremor characteristics. Here, we present the use of the iPhone® seismograph application (iSeismo by ObjectGraph LLC) in tremor detection and measurement. The freely available application uses the in-built accelerometer of the iPhone (or iPod Touch®) to measure movement in the X, Y, and Z axes relative to the device. These data are then displayed graphically showing the three axes and their predominant frequency band.
Although originally developed as an earthquake detector, here we show the iSeismo application can act as an easily accessible way for clinicians who own this popular device to measure pathological tremor simply and accurately.
The local research ethics committee approved methods in this study, and written informed consent was obtained from all patients. A total of 7 patients were used in this study, with the following conditions: Essential tremor (2 cases, 1 with implanted deep brain stimulator) Parkinson’s tremor, multiple sclerosis tremor, post-stroke tremor, dystonic tremor, and orthostatic tremor. This allowed comparison of the iPhone with the EMG analysis at various tremor frequencies.
The iPhone was strapped securely to the tremorous limb (either forearm or lower leg) and tremor was recorded using the iSeismo application for 30 s concurrently with EMG recordings. On-screen spectral analysis of tremor in three dimensions was saved as a screen capture for later comparison with the EMG.
Electromyograms (EMG) were recorded from disposable adhesive Ag/AgCl electrodes (H 207 PT; Kendall, Tyco International, Germany) placed roughly 1 cm apart on the belly of the muscle utilising a bipolar configuration. The muscles studied included flexor carpi radialis and extensor carpi radialis for all patients except orthostatic tremor where tibialis anterior was used. Parkinson’s, post-stroke, and dystonic tremors were captured at rest, while essential tremor and multiple sclerosis tremors were with arms straight out (postural position), and orthostatic tremor was with the patient standing. Signals were amplified (×1,000) using isolated CED 1902 amplifiers (Cambridge Electronic Design, Cambridge, UK), filtered (0–500 Hz) and digitised at 16 bit resolution using CED 1401 mark II analogue-digital converter (Cambridge Electronic Design, Cambridge, UK) at a sample rate of 2.5 kHz. Recordings were displayed online on a personal computer using Spike 2 (Cambridge Electronic Design, Cambridge, UK) and saved to disk for off-line analysis.
Data were analysed off-line using custom written software in the MatLab (Mathworks Inc., Natick, MA, USA) numerical simulation environment. A 30 s data segment free from large movement artefact, recorded during tremor, was exported at a sample rate of 1 kHz. Prior to further analysis data was bandpass filtered between 2 and 300 Hz. The signal was then full-wave rectified and the power spectral density (PSD) calculated using Welch’s modified periodogram approach utilising a Hanning window over 2 s disjointed sections [10]. Each PSD generated was evaluated for the frequency of tremor by identifying the highest amplitude peak and comparing this to the iPhone recording, which generates a frequency peak on-screen. The axis which best corresponded to the direction of movement was selected for this analysis. We focused only on the predominant frequency, as iSeismo does not properly measure amplitude, which allowed easy comparison across tremors and methods.
Section snippets
Results
The results of off-line EMG analysis matched well with the online iPhone tremor plots. Data are summarized in Table 1 and Fig. 1. Fig. 1 illustrates the compatibility of tremor assessment across these conditions. Individual plots for EMG and iPhone data are shown for the first case (essential tremor). For all subsequent cases, raw values were extracted and re-plotted on the same graph for direct comparison. Both methods produced similar bands of activity for the given tremor, and nearly
Discussion
The above cases demonstrate that the iPhone accelerometer, through the iSeismo interface, can be a useful way of quickly isolating the dominant tremor frequency in the clinical setting. This is perhaps the simplest method available for the quick assessment of tremor frequency, and as seen here can be applied to a variety of disease pathologies. All tremors were well detected by the iPhone and matched well with the EMG assessment. Even suppression of tremor with deep brain stimulation was
Acknowledgements
Oxford Functional Neurosurgery is supported by the Oxford Biomedical Research Centre of the UK National Institute of Health and Research, the Norman Collisson Foundation, educational grants from Medtronic Inc. and the Charles Wolfson Charitable Trust. R Joundi is supported by the Rhodes Trust. We thank Apple and ObjectGraph LLC for developing this software.
References (12)
Parkinsonism & related disorders
Tremor Parkinsonism Relat Disord
(2007)- et al.
The measurement of tremor using simple laser systems
J Neurosci Methods
(1994 Jul) - et al.
The contribution of tremor studies to diagnosis of parkinsonian and essential tremor: a statistical evaluation
J Clin Neurosci
(2002 May) - et al.
Consensus statement of the movement disorder society on tremor
Ad Hoc Scientific Committee Mov Disord
(1998) On certain tremors in organic cerebral lesions
Brain
(1904)- et al.
A new method for selective measurement of joint movement in hand tremor in Parkinson’s disease patients
J Med Eng Technol
(2007 Aug)
Cited by (69)
Analytical tools for evaluation pharmaceutical treatment of neurological diseases - A case study
2020, Procedia ManufacturingActivity-aware essential tremor evaluation using deep learning method based on acceleration data
2019, Parkinsonism and Related DisordersUpper limb intention tremor assessment: opportunities and challenges in wearable technology
2024, Journal of NeuroEngineering and RehabilitationQuantifying Movement Behavior of Chronic Low Back Pain Patients in Virtual Reality
2023, ACM Transactions on Computing for HealthcareDigital Phenotyping in Clinical Neurology
2022, Seminars in NeurologyMachine Learning Techniques for Automated Tremor Detection in the Presence of External Stressors
2022, International Journal of Circuits, Systems and Signal Processing
- ☆
The review of this paper was entirely handled by an Associate Editor, Jonathan Carr.