Treman: a tool for measuring tremor frequency from video sequences

Uhrikova Z., Ruzicka E., Hlavac V., Nugent C.: TremAn: A tool for measuring tremor frequency from video sequences, LETTERS TO Disorders, Vol. 25, No. 4, March 2010, pp. slow and fast as well as between apparently regular andirregular periodic movements. A more precise measure of the Segment 1. Patient exhibits choreiform movements dur- tremor frequency is provided by accelerometers1 and electro- ing the interview, which abruptly stop during performance of visual tasks. She can protrude her tongue for 10 attached to the patient’s body. Although wearable devices seconds. Her difficulty with tandem gait is suggestive of have been recently introduced integrating accelerometers and gyroscopes to capture movement features,3 these methods Author Roles: Fekete: Writing of the first draft and can still be viewed as being cumbersome and impractical for review and critique of manuscript and recording of initial routine clinical use. As an alternative, we propose a method video and editing; Jankovic: Review and critique of manu- of tremor frequency analysis based on video recordings of the patients—Tremor Analyzing Tool (TremAn)*.
TremAn measures the visible periodicity of the tremor.
This approach measures the changes in the image intensity (sum RGB components) of a selected area in the video sequence hence capturing movement of a specific body part in front of a background. From a theoretical perspective, Parkinson’s Disease Center and Movement Disorders Clinic the 3D movement of an object does not necessarily corre- spond to changes in image intensity. Our practical experi- ence tells us that changes in image intensity correspond well to tremor motion. The image intensity from theselected area, collected over time, forms a one-dimensionalperiodic signal. The frequency of this signal is measured based on its Fourier transform power spectrum computedwith the use of the implementation by Ooura.4 Higher lev- 1. Seitz P. Symbolism and organ choice in conversion reactions.
els of precision are reached by collecting the signal at spa- 2. Jankovic J. Treatment of hyperkinetic movement disorders. Lan- tial points, which are regularly distributed within the area of interest. The movement of the body part is expected to 3. Goetz CG, Chmura TA, Lanska DJ. History of chorea: part 3 of be consistent; hence, the spectra of all the selected points the MDS-sponsored history of movement disorders exhibit, Bar- are summed to produce one final spectrum. This approach celona, June 2000. Mov Disord 2001;16:331–338.
eliminates accidental noise, however, if two dominant fre- 4. Fahn S, Jankovic, J. Principles and practice of movement disor- quencies are present, two peaks can be seen in the resulting ders, First ed. Philadelphia: Elsevier; 2007. p 393–408.
power spectrum, with the higher peak taken as the principal 5. Thomas M, Jankovic, J. Psychogenic movement disorders: diag- nosis and management. CNS Drugs 2004;18:437–452.
TremAn allows for analysis of tremor recorded in most 6. Jankovic J, Ashizawa T. Tourettism associated with Huntington’s common video formats (.avi or .mpg). The assumptions made disease. Mov Disord 1995;10:103–105.
7. Jankovic J, Cloninger CR, Fahn S, Hallett M, Lang AE, Williams to ensure correct analysis are as follows: the tremor is visible DT. Therapeutic approaches to psychogenic movement disorders.
in the video, the area of interest is stable (the video sequence In: Hallett M, Fahn S, Jankovic J, Lang AE, Cloninger CR, Yudof- was captured with a fixed camera, with no shifting, zooming sky S, editors. Psychogenic Movement Disorders: Neurology and or focusing of the shot and the body part captured was not Neuropsychiatry. Philadelphia: AAN Enterprises and Lippincott moving markedly except for the tremor itself). The length of Williams and Wilkins; 2006. p 323–328.
the analyzed video sequence should be at least 5 seconds andthe sampling frequency should be at least 15 frames per sec-ond. To use the tool, users may simply open the videosequence in the application, select the area of interest (spe- cific body part) with a single mouse click in the video sequence, and then initiate the analysis.
The output of the algorithm is the frequency of tremor for the selected body part. The progress of the frequency in time is also recorded and can be used to investigate whether thefrequency was stable or changing. Several forms of visualiza-tion are offered: the signal progress, full frequency spectrum, Simple visual estimation of tremor frequency by a physi- or the frequency progress. These are all shown in graphs cian is a part of routine clinical examination in a patient suf- with the possibility to export the result as video, single fering from tremor, allowing a rough differentiation between image, or text file. Refer to Figure 1 for an example of theuser interface.
Refer to Video 1 for examples of tremor analysis. The first Additional Supporting Information may be found in the online segment shows a frequency measurement (3.22 6 0.2 Hz) of the right hand rest tremor in a patient with Parkinson’s dis- *The tool is available for free download at http://cmp.felk.cvut.cz/ ease. In the second patient with clinically probable psycho- Potential conflict of interest: Nothing to report.
genic tremor, the frequency of postural tremor of the left Published online 11 December 2009 in Wiley InterScience hand varies with time and decreases by more than 1 Hz in 25 (www.interscience.wiley.com). DOI: 10.1002/mds.22904 seconds. The third example analyses a recording of tremor FIG. 1. User interface of TremAn containing the selected video (the area of measurement is highlighted by the square). Three graphs displayingthe processed signal, the full frequency spectrum, and the principal frequency progress. Adjustable properties are listed on the right side.
taken from a previous publication in Movement Disorders the progress of the measured frequency, as mentioned in the Journal.5 In line with the results of electromyography men- text and shown in Figure 1. At the end of the processed tioned in the original article, the patient’s head tremor fre- sequence the example of the image output with the principal quency measured by TremAn is 5.3 Hz.
TremAn tool offers a convenient alternative to the com- Patient with probable psychogenic tremor.
mon methods of tremor analysis. The patient does not need Same features of video analysis as in Video segment 1.
to wear any recording technology. The tool can be also used Patient with Parkinson’s disease, with no-no to analyze archived videos from various sources.
head tremor (patient 5, segment 4, downloaded from Roze The results obtained by TremAn demonstrated a close cor- et al.).5 Same features of video analysis as in previous segments.
relation with tremor frequencies measured with accelerome-ters in the validation study with over 160 video sequences Acknowledgment: Czech Ministry of Health, project IGA (manuscript in preparation, partly presented at the recent IEEE EMBC Conference6). Also the analysis of the afore- MSM0021620849 and MSM684077003. Z. Uhrı´kova´ has mentioned archived video showed that the results correspond been supported by the European Commission under the well to tremor frequencies measured with electromyography.5 Marie-Curie research training network WARTHE MEST-CT- However, care should be taken that the tremor is recorded in standardized positions. For example, here, video segment 1 Author Roles: Uhrı´kova´: Conception and execution of the probably does not show as nice a peak as the others, because research project, execution of statistical analysis, writing of the tremor of the hand is superimposed on the underlying leg the first draft of manuscript; Ru˚zˇicˇka: Conception and organi- tremor. Measuring other features, such as amplitude of the zation of the research project, review and critique of statisti- tremor might be feasible only under very strict conditions cal analysis, review and critique of manuscript; Hlava´cˇ: Con- from calibrated videos and it is not practicable from ordinary ception and organization of the research project, design and review and critique of statistical analysis, review and critiqueof manuscript; Nugent: Conception of the research project,design and review and critique of statistical analysis, review Patient with Parkinson’s disease rest tremor of right hand. The area of measurement is highlighted in green. In the bottom of the video frame, the current time and measured frequency is displayed. Graphs 1–3 demonstrate the progress of the signal, the full frequency spectrum and A 29-year-old woman, with history of depression at the age of 18 and an abortion (fetal malformation) at age 24, was seen for acute onset of wave-like movements of the tongue with con-comitant cervical torsion (Video, Segment 1), 2 hours after she had taken 75 mg of venlafaxine prescribed by her general physi- cian due to complaints of sadness and anxiety. After a single 5- mg dose of biperiden, these movements subsided. In the follow- ing week, these symptoms recurred for three times and again remitted with 5 mg of biperiden. She remained asymptomaticfor 1 year. At this time, her general physician prescribed her 50mg of sertraline for depressive symptoms. Two hours after thefirst intake, an exuberant, irregular, arrhythmic, large-amplitude chin tremor developed (Video, Segment 2). This tremor per- sisted for several hours and was highly variable and responsive to both suggestion and placebo administration, the patient being remarkably indifferent to its occurrence. Both coactivation and entrainment signs could be elicited. Brain MRI, EEG, and athorough lab work-up were normal. Electromyography of mass-eter muscles fulfilled Milanov proposed criteria for psychogenic tremor.2 The diagnosis of psychogenic tremor could then be confidently made. Psychiatric evaluation elicited a conversion disorder. She is currently taking no medication. The frequency of the chin tremor episodes has diminished. The last occurred during child delivery and ceased spontaneously.
In the first episode, the movement disorder phenomenol- ogy, its clear-cut relation to drug administration, and itsprompt response to anticholinergic medication strongly sug- gest a venlafaxine-induced dystonia, similar to those previ-ously reported to occur with this antidepressant.3 1. Hallett M. Overview of human tremor physiology. Mov Disord In contrast, several features led us to label the second move- ment disorder as psychogenic. As previously described,1 the 2. O’Suilleabhain PE, Matsumoto JY. Time-frequency analysis of chin tremor of our patient was highly variable and both distracti- tremors. Brain 1998;121:2127–2134.
3. Giuffrida JP, Riley DE, Maddux BN, Heldman DA. Clinically bility and emotional indifference were remarkable. In addition, deployable Kinesia technology for automated tremor assessment.
two of the most specific signs in psychogenic tremor, coactiva- tion and entrainment, were found. Further supporting the diag- 4. Ooura T. General Purpose FFT. Available at: http://www.kurims.
nosis were symptom persistence after drug withdraw and the physiologic incongruent electromyographic pattern.
5. Roze E, Coeˆlho-Braga MC, Gayraud D, et al. Head tremor Of relevance is that citalopram-induced jaw tremor and several in Parkinson’s disease. Mov Disord 2006;21:1245–1248.
cases of bruxism due to SSRI drugs are well known, but with com- 6. Uhrı´kova´ Z, Sˇprdlı´k O, Hlava´cˇ V, Ru˚zˇicˇka E. Action tremor analysis plete symptomatic relief after drug withdraw.3,4 Furthermore, we from ordinary video sequence. Proceedings of the 31st Annual Inter- could not find any report of such a tremor induced by sertraline use.
national Conference of the IEEE Engineering in Medicine and Biol-ogy Society, Minneapolis, MN, USA, 2009. p 6123–6126.
In our patient, the development of a psychogenic movement disorder after an iatrogenic one is particularly noticeable, as ifshe had learned how it could arise. This is supported by datashowing that, in conversive patients, subsequent symptoms are rather similar to those of a past neurological disease, and as that, some kind of learning process appears to occur.1 Treatment of conversion disorder is rather complex. How- ever, making an accurate diagnosis and allowing the accep- tance of its psychogenic nature is essential for the treatmentto be successful and to avoid unnecessary investigations anddrug prescriptions.5 Psychogenic movement disorders are a huge diagnostic and therapeutic challenge for both neurologists and psychia- trists. A clear diagnostic is often difficult to establish, mainlybecause they can mimic a variety of neurological and psychi- Segment 1. The patient was seen in the emergency room atric disorders, or even concur with these.1 (ER) 2 hours after she has taken 75 mg of venlafaxine. Asthe video shows, she presented wave-like movements of thetongue and cervical torsion. These movements subsided with Additonal Supporting Information may be found in the online ver- 5 mg of biperiden and she was discharged home.
Segment 2. The patient presented to the ER with an exuberant, Potential conflict of interest: None reported.
Published online 11 December 2009 in Wiley InterScience atypical, and long-lasting chin tremor 2 hours after she has taken 50 (www.interscience.wiley.com). DOI: 10.1002/mds.22910 mg of sertraline. The unusual clinical picture demanded the admis-

Source: ftp://cmp.felk.cvut.cz/pub/cvl/articles/hlavac/UhrikovaMovementDisordersTramAn2010.pdf