a new technique for quantitative measurement of laryngeal videostroboscopic images ·...

A New Technique for QuantitativeMeasurement of LaryngealVideostroboscopic ImagesJoel A. Sercarz, MD; Gerald S. Berke, MD; David Arnstein, MD;Bruce Gerratt, PhD; Manuel Natividad

\s=b\The objective measurement of laryn-geal function and pathophysiology is one ofthe goals of current laryngeal research. Wedescribe a new computerized tool for voiceanalysis systems that allows the quantita-tive analysis of individual videostrobosco-pic images. We describe this new techniquecompared with previous methods of video-stroboscopic image analysis and discussits clinical and research applications.

(Arch Otolaryngol Head Neck Surg.1991;117:871-875)

Laryngologiste, speech patholo-gists, and voice researchers have

long sought to define objective mea¬sures of vocal function. Despite ad¬vances in the study of laryngeal func¬tion, surgeons typically rely on theirvisual examination of the larynx andtheir perception of the patient's vocal

quality in judging the results of thera¬py. Objective measures avoid relianceon the subjective interpretation of thelistener, which is prone to inconsisten¬cy, bias, and error. As phonosurgerybecomes more popular, objective anal¬ysis of the results may help to estab¬lish the benefits of new procedurescompared with those of standardtherapies.

Herein, we describe an addition toour office-based system for vocal anal¬ysis. The new system allows the quan¬titative analysis of individual video-stroboscopic frames. A central featureis a multipurpose software packagealso used to analyze satellite photo¬graphs. The system holds promiseboth for clinical laryngology and basiclaryngeal research.

Objective analysis of vocal functionmay include imaging and physiologicaerodynamic and acoustic measure¬ments.1 The information derived fromglottography can be further used togenerate indexes representing vocalcord motion.2 Videostroboscopy is an

imaging technique that provides reli¬able information about vocal fold

movement and overcomes the techni¬cal difficulties involved in high-speedmotion picture photography. Strobos-copy, however, has been limited by thelack of an accurate and readily avail¬able quantitative method of imageanalysis.

Laryngeal stroboscopy dates back to1878 and to Oertel, who shone a brightlight through a perforated rotatingdisk and used the alternating lightproduced to reflect off of a laryngealmirror. Because it was cumbersome,the device was received unenthusiasti¬cally.3 Kallen was the first to use themodern flash tube type of stroboscopein 1932.4 Recent advances in strobos¬copy units and video cameras, particu¬larly in fiberoptics, have allowed thewidespread use of videostroboscopyfor the analysis of laryngeal disorders.5

Stroboscopy circumvents the inabil¬ity of the eye to visualize very briefevents occurring in laryngeal vibra¬tion. An image on the human retinapersists for 0.2 second, and sequentialevents lasting milliseconds are not per¬ceptible. In electron stroboscopy, lightflashes can be synchronized with the

Accepted for publication January 8,1991.From the Divisions of Head and Neck Surgery,

UCLA Medical Center and West Los Angeles Vet-erans Administration Medical Center, and the Divi-sion of Head and Neck Surgery, Department ofSurgery, Stanford University Medical Center,Palo Alto, Calif.

Reprint requests to the Division of Head andNeck Surgery, UCLA Medical Center, Los Ange-les, CA 90024 (Dr Berke).

Downloaded From: http://jamanetwork.com/pdfaccess.ashx?url=/data/journals/otol/17372/ by a University of California - Los Angeles User on 05/30/2017

fundamental frequency of the subject'slaryngeal vibration. If synchronizedprecisely, a still picture is produced. Amoving image is simulated by produc¬ing flashes slightly out of phase withthe fundamental frequency of phona¬tion. The movement is created from a

montage of several different glotticcycles photographed at slightly differ¬ent points within the cycle.

Some knowledge of the physiology oflaryngeal vibration is necessary to cor¬

rectly interpret videostroboscopic im¬ages. Based on anatomic and vibratorystudies, Hirano and Kakita6 describedthe body/cover theory of vocal cordvibration, which emphasizes the im¬portance of a layered structure in vocalfold vibration. The stiff underlyingbody consists of the muscular portionof the vocal cords and the deep laminapropria. The pliable cover includes theoverlying mucosa as well as the super¬ficial lamina propria and accounts forthe mucosal wave visualized in laryn-gostroboscopy.6 The mucosal wave isbest visualized at loud, midrange pitchphonation and diminishes at the falset¬to register.3

In 1961, vonLeden7 described hisinitial clinical experience with strobos¬copy, emphasizing its use in the earlydetection of true vocal cord carcinomasand in the management of vocal nod¬ules. Mucosal lesions fixed to the un¬

derlying muscle, or body in the modelof Hirano and Kakita, abolish the mu¬cosal wave.36 Fex and Elmqvist8 andFex" subsequently noted the disap¬pearance of the traveling wave in pa¬tients with recurrent paralysis and itsearly reappearance with return of neu¬

ral function.Laryngostroboscopic images have

been analyzed perceptually with crite¬ria previously described by Kitzing10and Bless et al.11 The most frequentlyused variables are listed in Table 1 andwill be discussed briefly. Symmetrynot only includes the anatomic struc¬tures of the larynx but also reflectssymmetry of glottal opening and clos¬ing during each cycle. Glottic closurehas been divided into seven differentcategories by Bless et al; these labelsidentify adequacy of closure, gap loca¬tion (if present), and the presence ofbowing. Periodicity is assessed by syn¬chronizing the strobe flashes with the

Table 1.—Variables of a PerceptuallyBased Analysis of Videostroboscopy*SymmetryRegularityGlottic closure (complete vs incomplete)Amplitude of vibrationPresence of mucosal wave

Nonvibrating portionPresence of ventricular fold hyperadductionData are from Bless et al.'

frequency of phonation. If the larynxdoes not appear to be a still image,aperiodicity is present. This aperiodic-ity frequently is manifested clinicallyas hoarseness. The above measures

are not possible with indirect laryngos-copy, because in real time the freemargins of the vocal fold form a

blurred, vibrating mass.The measurement of the glottic area

has been a primary goal of prior objec¬tive studies of laryngeal imaging.High-speed laryngeal photographywas developed by the Bell Laborato¬ries in the 1930s. Early efforts to cal¬culate glottic area involved the use oftracing paper, on which laryngeal im¬ages from high-speed photographywere projected. Several laboratorieshave developed computerized methodsof analyzing projected laryngeal im¬ages.12,13 Tanabe et al13 employed a com¬

puter that analyzed a filmed strobeimage projected onto a data reductionsystem, a type of optical-to-digitaltransformation device. Their systemallowed measurement of the excursioncurves of the vocal folds and the calcu¬lation of glottal width, length, andarea.

A substantial drawback of suchmethods is the expense and cumber¬some nature of high-speed photogra¬phy compared with the current gener¬ation of laryngeal videostroboscopes.For example, typical filming of thelarynx would involve 10 seconds ofphonation or 50 000 frames, whichwould take weeks to analyze.

More recently, Koike and Imaizumi14used a data reduction-based systemsimilar to that of Tanabe et al13 tostudy stroboscopie images on 16-mmmotion picture film. Because of a lackof availability and ease, however, mo¬tion picture-based methods to calcu¬late the glottic area have not been

Table 2.—Cost of System for VoiceLaboratories Already Equipped With

Videostroboscopy

1990 Cost inEquipment_$US

Computer 2000Data translation (analogue 1500

to digital image board)Image Pro software 1000Mavigraph video printer 3000

widely used in clinical or research lar¬yngology. Herein, we summarize our

early experience with a new method ofquantitative analysis of videostrobos-copic images, outlining its possible re¬search and clinical applications.

MATERIALS AND METHODSOur office-based system for objective

voice evaluation has been previously de¬scribed.1" Briefly, the system involves a per¬sonal computer to digitize photoglottogra-phic and electroglottographic measures andacoustic input. Videostroboscopic recordingof vocal cord vibration has been used, but thevideo images were formerly analyzed basedon subjective criteria only.

The system used for objective measure¬ment ofvideostroboscopic images is shown inFig 1. The vocal cords were visualized on a

video monitor through a telescope (Naga-shima, Tokyo, Japan) (70°) that was connect¬ed to a charge-coupled device video camera

(IK-C30A, Tbshiba, Buffalo Grove, 111). A 50-mm lens (Bruel & Kjaer/Computar, Orange,Calif) and a diopter lens (Vivitar No. 1, SantaBarbara, Calif) were used to magnify andfocus the image of the glottis. Illuminationwas provided by a stroboscope (4914-4, Bruel& Kjaer) that employs a hand-held stetho¬scope for synchronization of the strobe pulsewith the vocal frequency. The videostrobos¬copic images were then recorded with a

three-fourths inch videotape recorder (VO9850, Sony) equipped with a time-code gen¬erator. A video time-date generator (WJ-810, Panasonic) was also used to mark eachvideo field.

Recorded vocalization of the vowel i was

played back with the job and shuttle searchmode of the videotape recorder. Desired vid¬eo fields were identified for digitization. Be¬fore digitization, the video field was stabi¬lized through the use of a video printer(Mavigraph UKP-5000, Sony, Teaneck, NJ).The image was then digitized with a videoFrame Grabber (2853-SQ, Data Translation,Marlboro, Mass) and stored in the computer.Thereafter, the images could be retrievedand observed on either the video monitor or a

printed photograph from the video printer.


Fig 1. —Diagram of system for objective analysis of videostroboscopic images. RGB indicates red,green, and blue; VCR, videotape recorder; and NTSC, National Television Standard Commission.

The area of the glottis was measured withmouse-driven software (Image Pro, MediaCybernetics, Silver Spring, Md). The area ofthe measured trace was expressed in pixelunits or calibrated to express the area inother units. Figure 2 illustrates the methodof calibration used for an in vivo canine ex¬

periment. A standard centimeter ruler was

lowered to the level of the glottis and mea¬

sured with the software, which gave a pixelvalue equivalent to 5 mm in length. To com¬

pare measurements from one subject, it iscritical that the camera is positioned at aconstant distance from the larynx.

The mouse-driven software can measure

the length or area of any portion of an imageas directed by the mouse. It also allows theuser to label images, rotate an image, mea¬

sure angles, combine two images, change thescale of an image, or place multiple videoframes on a single screen or photograph.

CLINICAL APPLICATIONS

Some uses of this system are pre¬sented in the following examples. Thefirst subject was a volunteer with anormal larynx; the second and thirdsubjects were patients with unilateralvocal cord paralysis. Patient 1 under¬went laryngoplasty, and patient 2 un¬derwent Isshiki type I thyroplasty.The surgical patients were examinedbefore and after the operation.

A still video image from the normalvolunteer is illustrated in Fig 3, top.This image was digitized and stored inthe computer. Figure 3, middle, showsthe digitized image after being recalledfrom the computer hard drive. As an

example of the use of the mouse-driven

Fig 2.—Calibration ruler placed at the level of thevocal cord for use in canine studies. The soft¬ware provides a pixel equivalent for a 5-mmlength.

software, we measured the relativearea of the vocal folds. The examina¬tion of the volunteer yielded normalfindings, and there was no history oflaryngeal disease. Figure 3, bottom,shows the outlining of the vocal foldwith the mouse. Five measurementswere made on five separate videoframes and averaged. The area of theright vocal cord averaged 11 735 pix¬els, and that of the left vocal cordaveraged 11 872 pixels. This result was

expected, given the symmetry of thelarynx and lack of disease. This tech¬nique has potential value for monitor¬ing vocal fold atrophy or reinnervationin the months following a laryngealparalysis.

Figure 4 demonstrates the comput¬erized analysis technique comparingrelative glottic area in patient 1 beforeand after laryngoplasty for left-sidedvocal cord paralysis, during the mostclosed and most open segments of thecycle. We have found that within a

montage of stroboscopie images, themost closed and most open portions ofthe cycle are easily identifiable. Figure4, top, illustrates the outlining of thetransglottic area with use of themouse-driven software. Area measure¬ments were made in patient 1 fromdigitized stroboscopie images afternormalization of the anterior to poste¬rior lengths. To normalize an image,the glottis is framed from its anteriorand posterior extent and placed into astandard-sized box with use of thesoftware. The images in Fig 4, bottom,have been normalized, allowing thecomparison of images obtained at dif-


Fig 3.—Top, A videostroboscopic image ob¬tained from a normal human volunteer. Middle,Same image after digitization by the videoFrame Grabber. Bottom, Same image showingoutlining of the vocal fold with use of the mouse-driven software.

ferent times with the camera at slight¬ly different distances from the larynx.

For the patient illustrated in Fig 4,bottom, the preoperative ratio of themost closed to most open areas was8327/9375 pixels, or 0.89, indicatingminimal glottic closure. Postoperative¬ly, the ratio of the most closed to themost open area was 815/4057 pixels, or

0.20. For comparison, the ratio is zero

in the normal case in which there is no

glottal gap during the most closed por¬tion of the cycle. The ratio of thepostoperative to preoperative mostclosed area is 815/8357 pixels, or 0.076,

Fig 4.—Videostroboscopic image obtained frompatient 1 with outlining of the glottis. Softwareprovides area in pixels. Bottom, Composite im¬age of patient 1 showing the most open andmost closed images before (preop) and after(postop) laryngoplasty. The patient had minimalglottic closure preoperatively.

indicating that the surgery reducedthe glottal gap during the most closedportion of the cycle to 7.6% of itspreoperative value. However, postop¬eratively there was still a glottal gapduring the most closed portion of thevocal cycle: the patient's most closedarea was still 20% of the most openarea.

Another potentially important mea¬sure is the symmetry of vocal cordexcursion, which can also be obtainedwith the digitized most open and mostclosed images. The images in Fig 5depict this procedure for patient 3who had right-sided vocal cord paresis.The thickness of the vocal fold for themost closed and most open images ismeasured on each side in pixels. Themaximal medial-to-lateral excursion islength a minus c (25.0 pixels) on thepatient's left side and length b minus d(8.0 pixels) on the right side, as shownin Fig 5. The symmetry ratio in patient3 is 25.0/8.0 pixels, or 3.1, indicatingthat the normal left cord's excursion is3.1 times greater than the weak rightcord's excursion. Such excursion asym-

Fig 5.—

Digitized videostroboscopic image of themost open phase of the cycle in patient 3 withright-sided vocal cord paresis showing measure¬ment of the symmetry ratio: a equals 53.0 pixelsand b equals 25.0 pixels. Measurement of thesymmetry ratio in patient 3 during the mostclosed phase of the cycle: c equals 28 pixels andd equals 17.0 pixels.

metry is an important factor of thepatient's voice disorder.

To compare ratios between patientswith either left- or right-sided paraly¬sis, we have arbitrarily determinedthat the vocal cord with greater excur¬sion is always placed in the numerator.A second variable is then used in casesof laryngeal paralysis to indicatewhether the normal "N" or paralyzed"P" cord demonstrates the greatestlateral excursion. For example, a ratioof 1.5N indicates that the normal corddemonstrates 1.5 times the excursionof the paralyzed cord.

COMMENT

The system described herein hastwo major clinical applications. First,the system allows quantitative analy¬sis of laryngeal variables, includingrelative glottic area, interarytenoiddistance, vocal cord area, and symme¬try of traveling wave motion. We havedemonstrated how these measures ap-


ply to patients with unilateral vocalcord paralysis before and after laryn¬goplasty. Second, the system perma¬nently archives digitized video images,thereby permanently storing the im¬ages on the disk memory of a relativelyinexpensive computer. This feature isimportant'for conditions of the larynxthat require the comparison of serialvideo images and eliminates time-con¬suming searches through videotape.Approximately six frames can be ar¬

chived on one high-density diskette,and up to nine images can be printedby the video printer.

Compared with previous methods ofobjective analysis of stroboscopie im¬ages,14 the methods described hereinhave the advantages of ease of opera¬tion, availability, and accuracy. Mostlaboratories lack the capability toquantify laryngeal variables throughhigh-speed photography, but centerswith stroboscopie units can readily ob-

tain the computer software and hard¬ware necessary for the system depict¬ed. Additional equipment costs forlaboratories with videostroboscopic ca¬

pability are listed in Table 2.Our group currently has only prelim¬

inary clinical experience with the sym¬metry ratio and the ratio of the mostclosed to most open portions of thecycle presented herein. However, anumber of studies comparing quantita¬tive videostroboscopic measures toacoustic, aerodynamic, and glottogra-phic measures are under way.

The technique is limited by the fun¬damental nature of both stroboscopyand the computer software. Strobos¬copy evaluates only a "snapshot" ofany single glottic cycle and is thereforenot suited to the assessment of pa¬tients with severe laryngeal aperiodic-ity, because extensive cycle-to-cyclevariations in vocal periodicity cannotbe evaluated with stroboscopy.3 In ad-

dition, the software system has twomain limitations. First, it is somewhatlabor-intensive, requiring a trained in¬dividual to carefully outline the laryn¬geal variable to be measured. Al¬though software for automatic analysisis not yet available, such automaticanalysis methods would save consider¬able time and expense. Second, theresolution of the videostroboscopic im¬ages following digitization is reducedbut still adequate, as shown in Figs 2and 3.

This system provides the opportuni¬ty to quantitatively analyze individualvideostroboscopic images. Further ad¬vances in videostroboscopic technologyand computer software promise to im¬prove the ease and accuracy of thesystem described herein.

This investigation was supported by grant DC00855-01 from the National Institutes of Health,Bethesda, Md, US Dept of Health and HumanServices.

References

1. Ward PH, Hanson DG, Gerratt BR, BerkeGS. Current and future horizons in laryngeal andvoice research. Ann Otol Rhinol Laryngol.1989;98:145-152.

2. Berke GS, Moore DS, Hantke DR, HansonDG, Gerratt BR, Burstein F. Laryngeal modeling:theoretical, in vitro, in vivo. Laryngoscope.1987;97:871-881.

3. Alberti PW. The diagnostic role of laryngealstroboscopy. Otolaryngol Clin North Am. 1978;11:347-354.

4. Padovan IF, Christman NT, Hamilton LH,Darling RJ. Indirect microlarygostroboscopy. La-ryngoscope. 1973;83:2035-2041.

5. Gould WJ, Kojima H, Lambiase A. A tech-nique for stroboscopic examination of the vocalfolds using fiberoptics. Arch Otolaryngol. 1979;105:285.

6. Hirano M, Kakita Y. Cover-body theory ofvocal fold vibration. In: Daniloff RG, ed. SpeechScience. San Diego, Calif: College-Hill Press;1985:1-46.

7. vonLeden H. The electronic synchron-strobo-scope: its value for the practicing laryngologist.Ann Otol Rhinol Laryngol. 1961;70:881-893.

8. Fex S, Elmqvist D. Endemic recurrent laryn-geal nerve paresis: correlation between EMG andstroboscopic findings. Acta Otolaryngol (Stockh).1973;75:368-369.

9. Fex S. Judging the movements of the vocalcords in larynx paralysis. Acta Otolaryngol(Stockh). 1970;263:82-83.

10. Kitzing P. Stroboscopy: a pertinent laryngo-logical examination. J Otolaryngol. 1985;14:151\x=req-\157.

11. Bless DM, Hirano M, Feder RJ. Videostro-

boscopic evaluation of the larynx. Ear Nose ThroatJ. 1987;66:48-58.

12. Hayden EH, Koike Y. A data processingscheme for frame by frame film analysis. FoliaPhoniatr (Basel). 1972;24:169-181.

13. Tanabe M, Kitajima K, Gould WJ, LambiaseA. Analysis of high-speed motion pictures of thevocal folds. Folia Phoniatr (Basel). 1975;27:77-87.

14. Koike Y, Imaizumi S. Objective evaluation oflaryngostroboscopic findings. In: Fujimura 0, ed.Vocal Physiology, Voice Production, Mechanismsand Functions. New York, NY: Raven Press;1988.

15. Berke GS, Hanson DG, Trapp TK, MooreDM, Gerratt BR, Natividad M. Office-based sys-tem for voice analysis. Arch Otolaryngol HeadNeck Surg. 1989;115:74-77.


a new technique for quantitative measurement of laryngeal videostroboscopic images ·...

Documents