Using a smart phone as a standalone platform for detection and monitoring of pathological tremors

Abstract

Introduction: Smart phones are becoming ubiquitous and their computing capabilities are ever increasing. Consequently, more attention is geared toward their potential use in research and medical settings. For instance, their built-in hardware can provide quantitative data for different movements. Therefore, the goal of the current study was to evaluate the capabilities of a standalone smart phone platform to characterize tremor.

Results: Algorithms for tremor recording and online analysis can be implemented within a smart phone. The smart phone provides reliable time- and frequency-domain tremor characteristics. The smart phone can also provide medically relevant tremor assessments.

Discussion: Smart phones have the potential to provide researchers and clinicians with quantitative short- and long-term tremor assessments that are currently not easily available.

Methods: A smart phone application for tremor quantification and online analysis was developed. Then, smart phone results were compared to those obtained simultaneously with a laboratory accelerometer. Finally, results from the smart phone were compared to clinical tremor assessments.

Keywords: Parkinson; essential tremor; long-term; movement disorder; telemedicine; telephone; tremor.

Figure 1

Example of tremor traces recorded with the smart phone and the accelerometer with their corresponding power spectrum. Top pane: example of a moderate amplitude tremor. Middle pane: example of a high amplitude tremor. Bottom pane: example of a low amplitude tremor (physiological tremor). (A) Example of a tremor trace recorded with the smart phone, (B) example of the tremor trace from the same trial as in (A) but recorded with the accelerometer, (C) power spectrum of the tremor trace recorded with the smart phone which was calculated with the algorithms implemented within the smart phone, (D) power spectrum of the tremor trace recorded with the smart phone which was calculated offline using our laboratory software, (E) power spectrum of the tremor trace recorded with the accelerometer which was calculated offline using our laboratory software.

Figure 2

Left column: correlation between smart phone data and accelerometer data for tremor amplitude, regularity, power distribution, median power frequency (MPF), and harmonic index (HI). Exact correlation coefficients and p values are in Table 2. Right column: Bland–Altman plots to evaluate the agreement between data from the smart phone and from the accelerometer for tremor amplitude, regularity, power distribution, median power frequency (MPF), and harmonic index (HI). Specific values for the Bland–Altman test (bias, SD) are shown in Table 2.

Figure 3

Comparison of tremor amplitude recorded with the smart phone according to the clinical tremor score each trial was given. Asterisk (*) indicates a significant difference from the previous group. p Values were 0.004, 0.002, 0.001, 0.001, and 0.017 for the paired comparisons 0–1, 1–2, 2–3, 3–4, and 4–5, respectively. The associated power for those tests was 0.788, 0.869, 0.973, 1.000, and 0.638, respectively.