Tremor Frequency Assessment by iPhone® Applications: Correlation with EMG Analysis

Abstract

Tremor frequency analysis is usually performed by EMG studies but accelerometers are progressively being more used. The iPhone® contains an accelerometer and many applications claim to be capable of measuring tremor frequency. We tested three applications in twenty-two patients with a diagnosis of PD, ET and Holmes’ tremor. EMG needle assessment as well as accelerometry was performed at the same time. There was very strong correlation (Pearson >0.8, p < 0.001) between the three applications, the EMG needle and the accelerometry. Our data suggests the apps LiftPulse®, iSeismometer® and Studymytremor® are a reliable alternative to the EMG for tremor frequency assessment.

INTRODUCTION

Tremor is defined as an involuntary rhythmic oscillation of a body part [1]. It is considered the most common hyperkinetic movement disorder, and it has a broad differential diagnosis, including Parkinson’s Disease (PD), Essential Tremor (ET), cerebellar/brainstem lesions. It is usually classified by body part affected, symmetry, if it’s present on rest, posture or action, and by its amplitude and frequency [2]. Tremor is prone to misdiagnosis: ET was erroneously diagnosed as PD or other tremor condition in 30–50% of the times [3]. Tremor frequency is a core diagnostic feature in a patient with tremor [1, 2]. While symmetry and body part affected can be easily evaluated by observation, frequency, calculated in Hertz (Hz), is more challenging. Even though physicians often use clinical rating scales based on observation alone, an objective assessment requires electromyographic or accelerometry measurements. This calls for burdensome equipment and trained operators, making it impractical to use in the clinic. Over the past years, other devices such as portable accelerometers have been used at the bedside [4]. Smartphones are now ubiquitous devices [5], and the iPhone® is particularly popular in our Neurology Department. The iPhone® has a built-in triaxial accelerometer which through an interface (a program or “application”) is capable of quantifying in Hz the frequency of a tremorous movement.

Previous studies have showed that smartphones’ built-in accelerometers are capable of detecting and analyzing tremorous movements [6], with results comparable to laboratory accelerometers and clinical evaluation [7, 8]. Two studies showed that signals obtained from smartphones combined with machine-learning techniques were capable of diagnosing a tremor as PD or not-PD [9] and DBS-ON andDBS-OFF in ET patients [10]. The majority of these studies used the iPhone® mainly as a signal detector for tremor and had to export the data to a central computer for analysis.

A paper by Joundi et al. (2011) showed that one iPhone® application (iSeismometer®), originally built for earthquake detection, could reliably assess tremor frequency, and values obtained by the app matched well with EMG analysis [4]. Senova et al. (2015) presented a new smartphone application (Itremor®) and showed that various tremor features (including frequency) correlated well with clinical evaluation in patients receiving deep brain stimulation (DBS) [11]. Recently, Kubben et al. also showed that a new smartphone app (TREMOR12) can also detect and analyze different tremor features [12]. Bhatti et al. demonstrated that two iPhone® applications (iSeismometer and LiftPulse®) were capable of accurately detecting high frequency tremor in lower limbs in patients with orthostatic tremor [13]. iSeismometer® and LiftPulse® are freely available apps for download in the AppStore®; Itremor® is only available from Ad Hoc distribution and not from the AppStore®. TREMOR12 is available in the AppStore® but requires data export or the installation of an extra processing module.

We aimed to test whether other apps available in the AppStore® that took advantage of the iPhone’s technology could reliably analyze tremor frequency in a clinical context.

There are several tremor assessment apps available in the App Store. Three applications (LiftPulse® – LP, iSeismometer® – iSeism, Studymytremor® – SMT) cited in the literature⁶ were studied. The values from these Apps and simultaneous EMG recordings were compared in movement disorders patients with tremor.

We show here that all three applications tested were similarly accurate in tremor frequency assessment and correlated very highly with EMG analysis.

MATERIAL AND METHODS

We examined 22 patients from our Movement Disorders outpatient clinic with a diagnosis of Parkinson’s Disease (n = 12, median age 67), Essential Tremor (n = 9, median age 65) and Holmes Tremor (n = 1, 72 year old patient with multiple sclerosis). The personal iPhone® of the first author was used (iPhone 5s). The iPhone was strapped to the relevant tremulous limb and the three applications were run consecutively in a random order. We used an adjustable armband to tightly secure the iPhone® around the wrist of the dorsal surface of the hand, carefully not to interfere with EMG recordings. PD patients’ tremor was assessed on the rest position and ET and Holmes tremor on posture. EMG needle assessment as well as accelerometry analysis were performed simultaneously on a Nicolet Viking Select® and Viking Select Tremor Master Software v8.3. The pair of muscles studied for each patient was selected on an individual basis according to the dominant tremor, and included pronator teres, brachioradialis, supinator, extensor digitorum communis, flexor pollicis longus, extensor carpi radialis, extensor pollicis longus, extensor pollicis brevis, flexor carpi radialis and extensor indicis.

Correlation between tremor frequency values of the three applications and the EMG needle and accelerometry was calculated (Pearson correlation). We then performed a linear regression and built a Bland-Altman graph for each pair application-EMG accelerometry [14].

RESULTS

Results of median tremor frequency by the apps and EMG needle and accelerometry for PD and ET patients were calculated, and a statistically significant difference between PD and ET patients was found for the three apps and for EMG (p < 0.002). Two patients’ tremor (both ET) were undetectable by EMG-needle due to low amplitude. Pearson correlation between the three applications and the EMG-needle and accelerometry was calculated (Table 1). Our results show strong correlations (Pearson >0.8, p < 0.001) between all the apps and the EMG (Table 1), as depicted in scatter plots (Fig. 1A-C). Bland-Altman graphs show that values are proximate throughout the spectrum of frequencies measured (Fig. 1), with σ_LP = 0.650, σ_iSeism = 0.789 and σ_SMT = 0.895).

DISCUSSION

These results show that, not only the previously studied iSeismometer®, but also the LiftPulse® and the Studymytremor® apps can reliably analyze tremor frequency at the bedside. The applications did not differ significantly from each other, but LiftPulse® showed a narrower confidence interval on the Bland-Altman graph and a higher correlation with the EMG-needle (0.928) and accelerometry (0.937) comparing with the other apps.

There are several tremor-detection applications available in the AppStore®. The majority of papers that test these applications refer to apps that are either unavailable [11], require data to be exported to a central computer or the installation of accessory programs [12], have not been compared to EMG [9–12], and were predominantly written from an engineering point of view. Joundi et al. showed that iSeismometer®’s tremor frequency analysis correlated with the EMG [4]. It was uncertain, however, whether other popular apps were as accurate and reliable as the iSeismometer®. Our results show that two other applications are just as accurate for tremor frequency evaluation. This is important because faulty programming as well as false advertising regarding commercially available apps may lead to erroneous conclusions with clear repercussions for patients.

We do not mean to suggest that the iPhone can replace the EMG altogether, but rather that the apps tested and the iPhone’s accelerometer seem to be reliable. Other useful tremor features, such as amplitude, were not taken into consideration in our study. Notably, two patients with essential tremor, with high frequency but very low amplitude postural tremors, could not adequately be assessed by EMG-needle, suggesting accelerometers could be advantageous for low amplitude tremors. The iPhone® demonstrated it can be very sensitive even in fine distal tremors when secured to the back of the patient’s hand. Additionally, even though we only tested upper limb tremors, we feel the iPhone® could also be used on the lower limbs. Bhatti and colleagues have showed that the iPhone® (particularly LiftPulse® and iSeismometer®) can be used as a screening tool for orthostatic tremor [13]. Bhatti et al. used a phlebotomy tourniquet; in our study, we used an adjustable armband to tightly secure the iPhone®, which could as easily be applied on the lower limbs. Joundi and Senova had also showed that the iSeismometer® demonstrated suppression of the tremor in one DBS patient very neatly[4, 11] – which supports the use of the iPhone® in other areas apart from diagnosis, such as monitoring treatment efficacy.

We conclude that LiftPulse®, iSeismometer® and Studymytremor® can act as a quick, accessible and reliable surrogate to tremor frequency analysis performed by the EMG. We consider that the portability and reliability of the iPhone® make a good case for its use in clinical practice and the possible integration of “tremor frequency” as an objective item in future (para)clinical rating scales.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.