Automated analysis of pen-on-paper spirals for tremor detection, quantification, and differentiation

Roopa Rajan; Reghu Anandapadmanabhan; Sharmila Nageswaran; Vineeth Radhakrishnan; Arti Saini; Syam Krishnan; Anu Gupta; Venugopalan Y Vishnu; Awadh K Pandit; Rajesh Kumar Singh; Divya M Radhakrishnan; Mamta Bhushan Singh; Rohit Bhatia; Achal Srivastava; Asha Kishore; MV Padma Srivastava

doi:10.4103/aomd.aomd_50_22

ORIGINAL ARTICLES

Year : 2023 | Volume : 6 | Issue : 1 | Page : 17-25

Automated analysis of pen-on-paper spirals for tremor detection, quantification, and differentiation

Roopa Rajan¹, Reghu Anandapadmanabhan¹, Sharmila Nageswaran², Vineeth Radhakrishnan³, Arti Saini¹, Syam Krishnan³, Anu Gupta¹, Venugopalan Y Vishnu¹, Awadh K Pandit¹, Rajesh Kumar Singh¹, Divya M Radhakrishnan¹, Mamta Bhushan Singh¹, Rohit Bhatia¹, Achal Srivastava¹, Asha Kishore⁴, MV Padma Srivastava¹
¹ Department of Neurology, All India Institute of Medical Sciences, New Delhi, India
² Department of Sensor and Biomedical Technology, VIT University, Vellore, Tamil Nadu, India
³ Department of Neurology, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Kerala, India
⁴ Department of Neurology, Sree Chitra Tirunal Institute for Medical Sciences and Technology, Kerala, India; Department of Neurology, Centre of Excellence in Neurosciences, Aster Medicity, Kochi, Kerala, India

Date of Submission	17-Oct-2022
Date of Decision	05-Dec-2022
Date of Acceptance	01-Jan-2023
Date of Web Publication	28-Apr-2023

Correspondence Address:
Roopa Rajan
Department of Neurology, All India Institute of Medical Sciences, New Delhi – 110029
India

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/aomd.aomd_50_22

Abstract

Objective: To develop an automated algorithm to detect, quantify, and differentiate between tremor using pen-on-paper spirals. Methods: Patients with essential tremor (n = 25), dystonic tremor (n = 25), Parkinson’s disease (n = 25), and healthy volunteers (HV, n = 25) drew free-hand spirals. The algorithm derived the mean deviation (MD) and tremor variability from scanned images. MD and tremor variability were compared with 1) the Bain and Findley scale, 2) the Fahn–Tolosa–Marin tremor rating scale (FTM–TRS), and 3) the peak power and total power of the accelerometer spectra. Inter and intra loop widths were computed to differentiate between the tremor. Results: MD was higher in the tremor group (48.9 ± 26.3) than in HV (26.4 ± 5.3; p < 0.001). The cut-off value of 30.3 had 80.9% sensitivity and 76.0% specificity for the detection of the tremor [area under the curve: 0.83; 95% confidence index (CI): 0.75, 0.91, p < 0.001]. MD correlated with the Bain and Findley ratings (rho = 0.491, p = 0 < 0.001), FTM–TRS part B (rho = 0.260, p = 0.032) and accelerometric measures of postural tremor (total power, rho = 0.366, p < 0.001; peak power, rho = 0.402, p < 0.001). Minimum Detectable Change was 19.9%. Inter loop width distinguished Parkinson’s disease spirals from dystonic tremor (p < 0.001, 95% CI: 54.6, 211.1), essential tremor (p = 0.003, 95% CI: 28.5, 184.9), or HV (p = 0.036, 95% CI: -160.4, -3.9). Conclusion: The automated analysis of pen-on-paper spirals generated robust variables to quantify the tremor and putative variables to distinguish them from each other. Significance: This technique maybe useful for epidemiological surveys and follow-up studies on tremor.

Keywords: Accelerometry, dystonic tremor, essential tremor, movement disorders, pen-on-paper spirals

How to cite this article:
Rajan R, Anandapadmanabhan R, Nageswaran S, Radhakrishnan V, Saini A, Krishnan S, Gupta A, Vishnu VY, Pandit AK, Kumar Singh R, Radhakrishnan DM, Bhushan Singh M, Bhatia R, Srivastava A, Kishore A, Padma Srivastava M V. Automated analysis of pen-on-paper spirals for tremor detection, quantification, and differentiation. Ann Mov Disord 2023;6:17-25

How to cite this URL:
Rajan R, Anandapadmanabhan R, Nageswaran S, Radhakrishnan V, Saini A, Krishnan S, Gupta A, Vishnu VY, Pandit AK, Kumar Singh R, Radhakrishnan DM, Bhushan Singh M, Bhatia R, Srivastava A, Kishore A, Padma Srivastava M V. Automated analysis of pen-on-paper spirals for tremor detection, quantification, and differentiation. Ann Mov Disord [serial online] 2023 [cited 2023 May 28];6:17-25. Available from: https://www.aomd.in/text.asp?2023/6/1/17/375302

Introduction

Tremor is a common neurological problem that causes considerable neurological disability and morbidity.^[1] Despite giant advances in medical diagnostic modalities, the diagnosis of diseases causing tremor and other movement disorders largely remains a clinical exercise. In 1998 and 2018, the consensus statements of the Movement Disorder Society on the classification of tremor are the current gold standard for its diagnosis.^[1],[2] Tremor severity is objectively rated using several validated clinical rating scales such as the Fahn–Tolosa–Marin tremor rating scale (FTM-TRS) and the essential tremor rating assessment scale (TETRAS), among others.^[3],[4],[5] Electrophysiological tremor analysis using a triaxial accelerometer combined with surface electromyography is often helpful in quantifying the frequency, amplitude, and other characteristics of a tremor.^[6] However, this is an investigative modality available at few movement disorder specialty centers, and it requires specialized equipment, trained professionals, substantial patient cooperation, and time investment. In addition, multiple variables including gravitational artefacts, effects of sensor positioning, and the inherent variability of the tremor often limit its widespread application in clinical practice.^[7] Therefore, a quick and reliable objective measurement of tremor is the need of the hour. This is essential not only in neurology clinics, but also during large-scale epidemiological surveys and other research settings, where often, inadequately trained individuals screen, enroll, and follow-up patients.^[8] In particular, this is relevant for studies assessing the prevalence or other characteristics of tremor disorders, including essential tremor (ET), dystonic tremor (DT), and Parkinson’s disease (PD).

In the clinic, asking the patient to draw a freehand Archimedes spiral using a pen on paper is part of the routine clinical examination of patients with tremor, specifically intended to identify action tremor of the upper extremity.^[9] The examining physician subjectively analyzes the hand-drawn spirals by visual inspection and estimates the severity of the tremor. The Bain and Findley spirography rating scale objectively rates the severity of the tremor from patient spirals, although it is limited by modest inter-rater agreement.^[10] Electronic surface input (tablet)-based methods may be used for objective quantification of the tremor from handwriting specimens or spirals.^{[11],[12],[13],[14]} For instance, an automated computer tremor score derived from spirals drawn on tablets correlated well with the clinical measures of the severity of tremor in ET and PD.^[15] Such digital spiral analysis may be a sensitive outcome measure for clinical trials of essential tremor^[16] and may be used to predict tremor^[17] or differentiate between them.^[18] Recently, artificial intelligence techniques were leveraged for the diagnosis of tremor from spiral images captured using a smartphone.^[19] While the digitizing tablets capture a host of signal variables for downstream analysis, the lack of time domain information and limited data capture, in particular, make the automated analysis of hand-drawn pen-on-paper spirals challenging. However, in resource-limited settings and for large epidemiological surveys, the use of minimal hardware and uncomplicated algorithms offers a distinct advantage. In this paper, we aimed to develop and validate an automated algorithm for the detection and quantification of tremor from hand-drawn pen-on-paper Archimedes spirals.

Methods

In this prospective study, we screened consecutive patients with upper extremity tremor and healthy volunteers at the Movement Disorder Clinic of the All India Institute of Medical Sciences, New Delhi, a tertiary-care university hospital. We included patients with the following Axis 1 diagnoses as per the 2018 Movement Disorder Society on the classification of tremors^[2]: essential tremor (ET, essential tremor and essential tremor plus), dystonic tremor (DT), and tremor combined with parkinsonism (PD).^[2] PD was diagnosed according to the UKPDS Brain Bank criteria.^[20] A movement disorder specialist neurologist, ascertained the diagnosis after clinical examination. Patients with alternative tremor diagnoses, those with no visible tremor in the dominant upper extremity, those with no visible tremor while on treatment, and those with severe tremor that prevented the ability to keep the pen on paper were excluded. Participants were examined for tremor at rest, on maintaining postures (posture 1: arms outstretched, elbows extended; posture 2: arms outstretched, elbows flexed with fingertips pointing at each other), and on action (finger–nose testing, writing, and pouring tasks). The Bain and Findley spiral rating scale was used to rate the severity of tremor from spiral drawings (median scores of 3 independent raters blinded to the diagnosis).^[10] A single trained investigator used the FTM–TRS scale to rate the severity of the tremor from structured videotaped neurological examinations.^[4] We used the Edinburgh Handedness Inventory to assess handedness.^[21] Limb tremor was recorded using a triaxial accelerometer attached to the middle phalanx of the index finger and surface electromyography (EMG) from the wrist flexors and extensors. Tremor was recorded from both the upper limbs at rest and while maintaining postures: posture 1 (arms outstretched forward, elbows extended) and posture 2 (arms abducted, elbows flexed with fingertips pointing towards each other). The following parameters were derived after Fourier transform and spectral analysis: 1) peak power (PP), 2) total power (TP) of the spectrum (1–30 Hz), and 3) peak frequency (PF).

Participants drew free-hand clockwise spirals with the dominant hand using a standard ball pen on white paper with a 15 × 15-cm square and a dot printed in the center. The drawing started from the dot, and the participants were not allowed to rest their elbow or wrist or take their hands off the paper while drawing. They were allowed a maximum of three attempts to produce an acceptable specimen. Spirals with severe tremor resulting in crisscrossing of the spirals or frequent interruptions were excluded. The spirals were scanned using a commercial scanner (resolution, 300 dpi) for offline storage. An in-house developed automated algorithm-calculated variables to quantify the tremor. [Figure 1]a shows the components of the algorithm and [Figure 1]b–e shows the representative images. In brief, the scanned spiral image underwent several pre-processing steps, including smoothening, filtering, and edge detection. Within the images, the boundaries of objects were identified using the Sobel edge detection algorithm. Line thickness was reduced to one pixel using erode function. The algorithm used the radius-angle transformation technique to unravel the spiral in the Cartesian coordinate system and convert it to a polar coordinate system, where the x-axis represented theta (angle) and the y-axis represented the distance (radius) from the center. Subsequently, it superimposed the best-fit line for an ideal Archimedes spiral on the radius–angle transformation of the input spiral [Figure 1]d. We subtracted the radius at each angle of the input Archimedes spiral from the best fit to identify the data offset and derive the total deviation of the spiral [Figure 1]e. We used the absolute values of data offset and computed their mean as a measure of total deviation from ideal spiral (mean deviation, MD) and the standard deviation as a measure of tremor variability (TV). In addition, we explored putative variables to differentiate among the tremor conditions. Mean inter and intra loop widths across four loops of the spiral were calculated from the spiral in the Cartesian coordinates. The movement disorder specialist who made the clinical diagnostic classification (reference standard) was not involved in spiral data acquisition or downstream processing. Diagnostic classification or other demographic information were not included in the automated algorithm that derived the spiral variables.

Figure 1: Flow chart of in-house algorithm (a) and representative images at various stages of processing, (b) original spiral, (c) pre-processed spiral, (d) best fit from ideal spiral, (e) data offset from best fit

Click here to view

Statistical analysis

The algorithm-derived parameters MD and TV correlated with the following: 1) TP and PP of the spectrum on tremor analysis, 2) the clinical rating scale for tremor (FTM–TRS total score), and 3) the clinical rating scale for spirals (the Bain and Findley spiral score). Spearman’s rho was used to assess the correlation between the spiral analysis parameters and tremor analysis parameters or clinical rating scales. The accelerometer data were log transformed as they deviated from the normal distribution. Receiver operating characteristic curve statistics were applied to identify cut-offs to detect the tremor spirals from controls. Mixed ANOVA with a repeated measures design and post-hoc Tukey test were used to compare loop widths among the different diagnostic groups. A P value of 0.05 was considered to be statistically significant. Statistical analysis was conducted using SPSS 20.0 (IBM Corp. Armonk, NY).^[22]

Sample size estimation

As this was an exploratory study using novel spiral-derived parameters, formal sample size estimation was not possible a priori. A previous study showed that approximately 25 healthy volunteers are required to generate laboratory-specific normative data for accelerometery–EMG tremor analysis.^[23] Extrapolating the data obtained for initial validation of our technique, we aimed to recruit 100 participants (25 ET, 25 DT, 25 PD, and 25 healthy volunteers).

Minimum detectable change analysis

Test–retest analysis was performed on 14 participants to compute the minimum detectable change (MDC) value. We included patients with DT who were part of a randomized clinical trial of botulinum neurotoxin versus placebo.^[24] Spiral drawings from 14 participants who received placebo injections as part of the double-blinded study were utilized to compute MDC after the completion of the trial. We used spirals drawn at baseline and after 6 weeks of placebo injection. MDC was calculated using the following formula: MDC = 1.96 × standard deviation (SD) and expressed as MDC% = MDC*100/Mean_baseline.^[25],[26]

Standard protocol approvals, registrations, and patient consents

This study was approved by the Institute Ethics Committee, All India Institute of Medical Sciences, New Delhi (IEC-173/07.04.2017). Written informed consent was obtained from all participants. The study was registered in the Clinical Trials Registry of India (http://ctri.nic.in; CTRI/2017/09/009766).

Data availability

Deidentified data will be made available by the corresponding author on request.

Results

During September 2017 and March 2020, we screened 440 patients with upper limb tremor and/or parkinsonism and recruited 75 patients (25 ET, 25 DT, 25 PD) and 25 healthy volunteers [Figure 2]. Mean (SD) age of the study population was 44.8 (16.8) years, and 86% were male. The mean duration of symptoms in patients was 8.0 years (7.0). The baseline characteristics are shown in [Table 1]. No adverse events related to the study procedures were noted.

Figure 2: Patient flow diagram
MD, mean deviation; TV, tremor variability

Click here to view

Table 1: Baseline characteristics

Click here to view

Spiral variables for detection of tremor

Spiral images from seven participants (ET, two; PD, five) were excluded from the radius–angle transformation analysis due to frequent interruptions or crisscrossing. Mean MD values were higher in patients with tremor compared to healthy volunteers (tremor, MD = 48.9 ± 26.3; HV, MD = 26.4 ± 5.3; p < 0.001; [Figure 3]a). The mean TV values were not different between patients and healthy volunteers (tremor, TV = 70.8 ± 10.1; HV, TV = 74.6 ± 8.0; p = 0.070; [Figure 3]b). The receiver operating characteristic curve analysis showed that MD values >30.3 had sensitivity of 80.9% and specificity of 76.0% to detect tremor [area under the curve: 0.83; 95% confidence index (CI), 0.75–0.91; p < 0.001 [Supplementary Figure 1]]. TV at a cut-off value of 73.1 had sensitivity of 58.8% and specificity of 60.0% to detect tremor (area under the curve: 0.61; 95% CI, 0.49–0.74; p = 0.093 [Supplementary Figure 2]). MD was the better parameter to detect tremor.

Figure 3: Box-plot showing MD (a) and TV (b) values in patients with tremor and healthy volunteers

Click here to view

Supplementary Figure 1: Receiver operating curve curve showing mean deviation between patients and controls

Click here to view

Supplementary Figure 2: Receiver operating curve curve showing tremor variability between patients and controls

Click here to view

Spiral variables for tremor quantification

The spiral-derived variable MD showed good correlation with the Bain and Findley spiral ratings (Spearman’s correlation: MD, rho = 0.491, p = 0 < 0.001). MD correlated well with the accelerometric measures of postural tremor (Spearman’s correlation for posture 1: TP, rho = 0.366, p < 0.001 and PP, rho = 0.402, p < 0.001; posture 2: TP, rho = 0.338, p = 0.001 and PP, rho = 0.387, p < 0.001). MD did not correspond with the accelerometric measures of rest tremor (Spearman’s correlation for rest tremor: TP, rho = 0.107, p = 0.312 and for PP, rho = 0.155, p = 0.138). In the tremor group, MD values increased with increasing FTM–TRS part B scores (rho = 0.260, p = 0.032). There was no correlation with the FTM–TRS total (rho = 0.159, p = 0.194), part A (rho = 0.147, p = 0.231), or part C (rho = -0.003, p = 0.983) scores. The minimum detectable change for spiral analysis using the MD values was computed to be 19.9% [Supplementary Table 1].

Supplementary Table S1: Correlation of the MD of the spiral variable with clinical rating scales and electrophysiological measures of tremor

Click here to view

The second spiral derived variable, TV, did not correlate with the Bain and Findley spiral score (Spearman’s correlation: TV, rho = 0.205, p = 0.093). There was no correlation between TV and the accelerometric measures of tremor frequency at rest (Spearman’s correlation for PF: rho = 0.457, p = 0.078), posture 1 (Spearman’s correlation for PF: rho = -0.085, p = 0.417), or posture 2 (Spearman’s correlation for PF: rho = -0.201, p = 0.054). There was no correlation between TV and FTM–TRS total, part A, part B, or part C scores (Spearman’s correlation for FTM–TRS total: rho = -0.045, p = 0.715; part A: rho = -0.109, p = 0.375; part B: rho = 0.084, p = 0.497; and part C: rho = -0.069, p = 0.578 [Supplementary Table 2]).

Supplementary Table S2: Correlation of the TV of the spiral variable with clinical rating scales and electrophysiological measures of tremor

Click here to view

Spiral variables to distinguish between tremor syndromes

There were no differences in MD and TV values between the different tremor groups [Table 2]. However, the loop widths significantly differed among the participant groups. Mean intra loop widths were lesser in PD compared to DT (p < 0.001, 95% CI: 54.59–211.06), ET (p = 0.003, 95% CI: 28.46–184.93), or HV (p = 0.036, 95% CI: -160.37–-3.90; [Figure 4]a). In addition, inter loop widths were lesser and decreasing in PD compared to DT (p < 0.001, 95% CI: 68.24–263.82), ET (p = 0.003, 95% CI: 35.58–231.16), or HV (p = 0.036, 95% CI: -200.46–-4.88; [Figure 4]b).

Table 2: Algorithm-derived variables for tremor classification

Click here to view

Figure 4: Loop widths to distinguish PD from ET, DT, and HV; (a) Intra and (b) interloop widths are markedly lower in PD compared to those in ET, DT, and HV
DT, dystonic tremor; ET, essential tremor; PD, Parkinson’s disease; HV, healthy volunteers

Click here to view

Discussion

We developed an automated analysis algorithm for the objective analysis of hand-drawn pen-on-paper spirals and validated it with clinical rating scales and triaxial accelerometry-surface electromyography tremor analysis for the detection and quantification of tremor. Spiral variable MD successfully differentiated healthy subjects from patients with tremor. In addition, we explored putative variables such as intra and inter loop width, which may be useful to distinguish between tremor syndromes. Loop widths were helpful in differentiating PD tremor from ET and DT. Our findings suggest that this analysis technique can be used in the clinic and other nonspecialized settings for the detection and quantification of tremor. Furthermore, this technique is useful to quantify the response to therapeutic interventions at regular intervals over the course of the treatment.

The analysis algorithm used in this study is broadly similar to previously described techniques that utilize radius–angle transformation and computation of the deviation from the baseline.^[27] However, distinct from previous efforts that attempted to derive an absolute tremor amplitude from spirals,^[28] we derived variables that are surrogate measures of tremor amplitude, without attempting to calculate the actual tremor amplitudes from this data. Good correlation with clinical spiral ratings, tremor rating scale scores, and log transformed power of the spectrum from the accelerometry data show that the derived variable MD is a robust measure of tremor severity. Marked difference in the MD values between the patients with tremor and control subjects translated to a cut-off value with reasonable sensitivity and specificity for the detection of tremor. In addition, consistent with previous studies, our results suggest that spiral analysis is an accurate measure of action/postural tremor, and it may not provide detailed information regarding rest tremor.^[11]

Previous studies on automated analysis of pen-on-paper spirals are listed in [Table 3]. In addition to the usefulness of these reports, which highlighted several approaches to pen-on-paper spiral analysis, our study supplements the data on the radius–angle transformation technique by computing two novel variables and prospectively validating the results beyond the Bain and Findley classification, as reported by Kraus et al.^[27] While the Bain and Findley scale is a validated scale for visual spiral rating, the inter rater agreement is modest; therefore, we used FTM–TRS and tremor accelerometry for additional validation of the technique. Wille et al.^[29] used a histogram-based technique on 109 spiral drawings from patients with uncharacterized tremor and showed that tremor severity could be binned into four categories, similar to clinical scales. Our approach generates a tremor severity variable on a continuous scale, without binning into distinct categories. The sensitivity (80.9%) and specificity (76.3%) of our system to detect tremor is similar to a recent report on diagnostic accuracy with an artificial intelligence algorithm (79% for normal controls).^[19]

Table 3: Studies reporting the analysis of pen-on-paper spirals

Click here to view

The MDC of a diagnostic assay or transducer reflects its sensitivity to change. If MDC is higher than the natural variability of the measured signal, the sensitivity to detect meaningful changes may be limited, as observed with tremor accelerometry. The MDC for this system is lower compared to that previously reported for tablet-based systems or clinical rating scales (51–67%).^[25] This may be due to the elimination of additional sources of variability that occur while using a digitizing tablet, such as speed and pressure of writing, which are not captured by the pen-on-paper system. In particular, a lower MDC is useful for repeated measurements, such as in clinical trials, to detect serial change in tremor amplitudes after an intervention. Although the lack of time domain information made the analysis challenging, especially in terms of frequency analysis, clinically meaningful variables may be derived in its absence.

In terms of distinction between tremor syndromes, the putative variables we explored (intra and inter loop widths) show promise in distinguishing PD spirals from others. This is an objective way of quantifying the decreasing spiral widths in PD from the increasing or static widths in DT or ET. However, there were no remarkable differences between the spiral widths in DT and ET, suggesting that the tremor characteristics may be similar in these two syndromes. To the best of our knowledge, is a unique aspect of our study as previous studies have not attempted to distinguish among tremor syndromes (except Ishii et al.,^[19] who distinguished ET from cerebellar disease) and were focused on the measurement of tremor severity. As the variables were calculated directly from the spiral in Cartesian coordinates, it could apply to all spirals, irrespective of crisscrossing or interruptions. Although statistically significant differences were identified, the confidence intervals for the loop width variables were wide Furthermore, a larger sample size may have to be studied to validate these findings; therefore, we consider them putative at present.

Strengths of the study

A major strength of this study is the prospective validation of our technique against two clinical rating scales and accelerometry–EMG tremor analysis in a large cohort of participants. In our study, the spirals were drawn by patients with three major tremor disorders and healthy volunteers, while other studies have used limited scales or a smaller number of participants. As shown in [Table 3], limited approaches were reported to analyze pen-on-paper spirals compared to digitally acquired spirals, which leverage the availability of time domain information and multiple variables such as acceleration, velocity, and pen pressure that may be acquired using a digitizing tablet. The lack of the time domain information and other variables makes the analysis of simple pen-on-paper spirals challenging. However, this maybe a distinct advantage in certain clinical situations. For instance, epidemiological studies have used pen-on-paper spirals and handwriting specimens to screen for tremor in several subjects, both in high- and low-resource settings.^[32],[33] Such screening has traditionally been performed by individual raters, which includes an unavoidable element of subjectivity and problems with the inter-rater agreement. An automated method to detect and quantify tremor from pen-on-paper spirals can make such observations more objective, while maintaining the scalability of a pen-on-paper test. An additional strength of our algorithm is its MDC, which is lower than that reported for digitized input systems. In particular, lower MDC makes the system suitable to detect smaller changes on serial measurements. In addition, it increases its sensitivity as an outcome measure for clinical trials.

Limitations of the study

Despite our promising findings, there are certain limitations to our study. The technique-related limitations include a prominent ceiling effect, since for all spiral-based measures of tremor, the system cannot be used by patients with severe tremor that precludes the ability of patients to keep a pen on paper at all. In addition, frequent crisscrossing and interruptions in drawing may lead to unreliable results and erroneous unravelling of the spiral; therefore, we had to exclude five PD spirals from the radius–angle transformation analysis. However, the loop width variables were not affected by this issue and merit further exploration as measures of tremor severity. Many variables reported to be useful in distinguishing tremor syndromes, such as a predominant tremor axis in ET, are not easily derived by conventional algorithms such as ours. This maybe an area where artificial intelligence-based algorithms may be of incremental value. In terms of the study design, external validation of our results in a larger multicentric cohort is currently lacking.

Conclusion

Our results suggest that clinically meaningful and methodologically robust variables can be derived using pen-on-paper spirals with minimal hardware or software requirements. This may be a potentially useful tool not only in resource-limited settings, but also in general neurology clinics and large-scale epidemiological surveys for the detection and quantification of tremor.

Acknowledgements

We are thankful to all the subjects who have participated in this study.

Author contributions

Research project:

Conception: RR

Organization: RR, RA, SN, VR, SK

Execution: RR, RA, SN, VR, AS, SK

Statistical analysis:

Design: RR, RA

Execution: RR, RA

Review and Critique:

Manuscript preparation: RR, RA, SN, VR, AS, SK, AG, VYV, AKP, RKS, DMR, MBS, RB, AS, AK, MVPS

Writing of the first draft: RR, RA, SN, VR, AS, SK

Review and Critique. RR, RA

Ethical compliance statement

This study was approved by the Institute Ethics Committee, All India Institute of Medical Sciences, New Delhi (IEC-173/07.04.2017). Written informed consent was obtained from all participants. The study was registered in the Clinical Trials Registry of India (http://ctri.nic.in; CTRI/2017/09/009766).

Financial support and sponsorship

DST-SERB Early Career Research Award (ECR/2016/001862) to Roopa Rajan.

Conflicts of interest

There are no conflicts of interest.

References

1.	Deuschl G, Bain P, Brin M; Ad Hoc Scientific Committee. Consensus statement of the movement disorder society on tremor. Mov Disord 1998;13(Suppl 3):2–23.
2.	Bhatia KP, Bain P, Bajaj N, Elble RJ, Hallett M, Louis ED, et al. Consensus statement on the classification of tremors. From the task force on tremor of the International Parkinson and Movement Disorder Society. Mov Disord 2018;33:75–87.
3.	Fahn S, Tolosa E, Marin C Clinical rating scale for tremor. In: Parkinson’s Disease and Movement Disorders. 1988. p. 225–34.
4.	Fahn S, Tolosa E, Marin C Clinical rating scale for tremor. In: Parkinson’s Disease and Movement Disorders. 2nd ed. 1993. p. 225–34.
5.	Elble RJ The essential tremor rating assessment scale. J Neurol Neuromed 2016;34–8.
6.	Haubenberger D, Abbruzzese G, Bain PG, Bajaj N, Benito-León J, Bhatia KP, et al. Transducer-based evaluation of tremor. Mov Disord 2016;31:1327–36.
7.	Elble RJ Gravitational artifact in accelerometric measurements of tremor. Clin Neurophysiol 2005;116:1638–43.
8.	Louis ED Utility of the hand-drawn spiral as a tool in clinical-epidemiological research on essential tremor: Data from four essential tremor cohorts. Neuroepidemiology 2015;44:45–50.
9.	Alty J, Cosgrove J, Thorpe D, Kempster P How to use pen and paper tasks to aid tremor diagnosis in the clinic. Pract Neurol 2017;17:456–63.
10.	Bain PG, Findley LJ Assessing Tremor Severity: A Clinical Handbook. London: Smith-Gordon; 1993.
11.	Pullman SL Spiral analysis: A new technique for measuring tremor with a digitizing tablet. Mov Disord 1998;13(Suppl 3):85–9.
12.	Elble RJ, Brilliant M, Leffler K, Higgins C Quantification of essential tremor in writing and drawing. Mov Disord 1996;11:70–8.
13.	Legrand AP, Rivals I, Richard A, Apartis E, Roze E, Vidailhet M, et al. New insight in spiral drawing analysis methods - Application to action tremor quantification. Clin Neurophysiol 2017;128:1823–34.
14.	Sisti JA, Christophe B, Seville AR, Garton ALA, Gupta VP, Bandin AJ, et al. Computerized spiral analysis using the iPad. J Neurosci Methods 2017;275:50–4.
15.	Rudzińska M, Izworski A, Banaszkiewicz K, Bukowczan S, Marona M, Szczudlik A Quantitative tremor measurement with the computerized analysis of spiral drawing. Neurol Neurochir Pol 2007;41:510–6.
16.	Haubenberger D, Kalowitz D, Nahab FB, Toro C, Ippolito D, Luckenbaugh DA, et al. Validation of digital spiral analysis as outcome parameter for clinical trials in essential tremor. Mov Disord 2011;26:2073–80.
17.	DelMastro H, Ruiz J, Gromisch E, Garbalosa J, Triche E, Olson K, et al. Quantification characteristics of digital spiral analysis for understanding the relationship among tremor and clinical measures in persons with multiple sclerosis. J Neurosci Methods 2018;307:254–9.
18.	Yu NY, Van Gemmert AWA, Chang SH Characterization of graphomotor functions in individuals with Parkinson’s disease and essential tremor. Behav Res Methods 2017;49:913–22.
19.	Ishii N, Mochizuki Y, Shiomi K, Nakazato M, Mochizuki H Spiral drawing: Quantitative analysis and artificial-intelligence-based diagnosis using a smartphone. J Neurol Sci 2020;411:116723.
20.	Hughes AJ, Daniel SE, Kilford L, Lees AJ Accuracy of clinical diagnosis of idiopathic Parkinson’s disease: A clinico-pathological study of 100 cases. J Neurol Neurosurg Psychiatry 1992;55:181–4.
21.	Oldfield RC The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia 1971;9:97–113.
22.	IBM Corp. IBM SPSS Statistics for Windows. Version 20.0.
23.	Raethjen J, Lauk M, Köster B, Fietzek U, Friege L, Timmer J, et al. Tremor analysis in two normal cohorts. Clin Neurophysiol 2004;115:2151–6.
24.	Rajan R, Srivastava AK, Anandapadmanabhan R, Saini A, Upadhyay A, Gupta A, et al. Assessment of botulinum neurotoxin injection for dystonic hand tremor: A randomized clinical trial. JAMA Neurol 2021;78:302–11.
25.	Elble RJ, Ellenbogen A Digitizing tablet and Fahn-Tolosa-Marín Ratings of Archimedes spirals have comparable minimum detectable change in essential tremor. Tremor Other Hyperkinet Mov (N Y) 2017;7:481.
26.	Spooner J, Steven AD, Donald WM Minimum detectable change analysis. Tech Notes 2011; Developed for U.S. Environmental Protection Agency.
27.	Kraus PH, Hoffmann A Spiralometry: Computerized assessment of tremor amplitude on the basis of spiral drawing. Mov Disord 2010;25:2164–70.
28.	Surangsrirat D, Intarapanich A, Thanawattano C, Bhidayasiri R, Petchrutchatachart S, Anan C Tremor assessment using spiral analysis in time-frequency domain. In: 2013 Proceedings of IEEE Southeastcon. 2013. p. 1–6.
29.	Wille A, Sangaré M, Winter S Analysis of patterns in tremor diagnosis spiral drawings for automated classification. Biomed Tech (Berl) 2013;58(Suppl 1):/j/bmte.2013.58.issue-s1-L/bmt-2013-4278/bmt-2013-4278.xml. doi: 10.1515/bmt-2013-4278.
30.	Miralles F, Tarongí S, Espino A Quantification of the drawing of an Archimedes spiral through the analysis of its digitized picture. J Neurosci Methods 2006;152:18–31.
31.	Galli M, Vimercati SL, Manetti E, Cimolin V, Albertini G, De Pandis MF Spiral analysis in subjects with Parkinson’s disease before and after levodopa treatment: A new protocol with stereophotogrammetric systems. J Appl Biomater Funct Mater 2014;12:e107-12.
32.	Louis ED, Hafeman D, Parvez F, Alcalay RN, Islam T, Siddique AB, et al. Prevalence of essential tremor in Araihazar, Bangladesh: A population-based study. Neuroepidemiology 2011;36:71–6.
33.	Louis ED, Thawani SP, Andrews HF Prevalence of essential tremor in a multiethnic, community-based study in Northern Manhattan, New York, N.Y. Neuroepidemiology 2009;32:208–14.