Measurement of Hyoid and Laryngeal Displacement in Video Fluoroscopic Swallowing Studies: Variability, Reliability, and Measurement Error

Transcription

1 Dysphagia DOI /s ORIGINAL ARTICLE Measurement of Hyoid and Laryngeal Displacement in Video Fluoroscopic Swallowing Studies: Variability, Reliability, and Measurement Error Isaac Sia Pamela Carvajal Giselle D. Carnaby-Mann Michael A. Crary Received: 29 April 2011 / Accepted: 24 June 2011 Ó Springer Science+Business Media, LLC 2011 Abstract Video fluoroscopy is commonly used in the study of swallowing kinematics. However, various procedures used in linear measurements obtained from video fluoroscopy may contribute to increased variability or measurement error. This study evaluated the influence of calibration referent and image rotation on measurement variability for hyoid and laryngeal displacement during swallowing. Inter- and intrarater reliabilities were also estimated for hyoid and laryngeal displacement measurements across conditions. The use of different calibration referents did not contribute significantly to variability in measures of hyoid and laryngeal displacement but image rotation affected horizontal measures for both structures. Inter- and intrarater reliabilities were high. Using the 95% confidence interval as the error index, measurement error was estimated to range from 2.48 to 3.06 mm. These results address procedural decisions for measuring hyoid and laryngeal displacement in video fluoroscopic swallowing studies. Keywords Hyoid and laryngeal displacement Video fluoroscopy Measurement variability Measurement error Swallowing Deglutition Deglutition disorders I. Sia P. Carvajal M. A. Crary (&) Department of Speech, Language, and Hearing Sciences, College of Public Health and Health Professions, University of Florida, P. O. Box , Gainesville, FL , USA mcrary@phhp.ufl.edu G. D. Carnaby-Mann Department of Behavioral Science and Community Health, College of Public Health and Health Professions, University of Florida, 101 South Newell Drive, Gainesville, FL 32611, USA The study of swallowing kinematics has typically included movement of the hyoid bone and larynx because of the relative ease of tracking these structures and their perceived functional importance in swallowing [1, 2]. Although video fluoroscopy is commonly used for the analysis of hyoid and laryngeal kinematics, radiographic magnification and radial distortion pose inherent limitations to the accuracy of this procedure [3]. Radiographic magnification causes the distance between any two points on the image to appear larger, while radial distortion causes an image to be increasingly stretched toward the periphery [4]. As a result of these radiographic influences, a necessary step in kinematic analyses is image calibration with referent objects of known dimensions. Across published studies, the type and location of these referents vary, including under the chin, on the lateral neck, or even on the fluoroscope away from the subject s body [2, 5 11]. In addition, in the absence of a calibration referent of known dimensions, some researchers have used the third cervical vertebra (C3) as a calibration referent. For example, Kim and McCullough [11] and Pauloski et al. [12] assigned the anterior edge of C3 a length of 15 mm based on skeletal and radiological references. Variability among calibration referents is likely to introduce variability among measured distances. This variability may well introduce interpretative error both within and across investigations. Despite this overt possibility, no study to date has examined measurement variability associated with different calibration referents. In addition to radiographic image variables, a change in head position during a swallow has the potential to affect measurement accuracy. To adjust for differences in head position, investigators may employ a vertically aligned axis of reference from which to obtain hyoid and laryngeal displacement measures. A common technique is to rotate

2 the radiographic image to vertically align the anterior edges of the second to fourth cervical vertebra, C2 C4 [2, 9, 11, 13]. Nevertheless, not all investigators adjust for changes in head position during or between swallows, relying instead on a single point of reference [13 16]. Furthermore, a variety of different reference axes have been utilized in literature [6, 8, 10, 17, 18]. According to Zu et al. [19], the application of even slightly different reference axes may affect resulting measurements. Thus, though the vertically aligned C2 C4 reference axis is commonly used in measurement of hyoid and laryngeal displacement, its impact on measurement variability has not been described. This study examined measurement variability in hyoid and laryngeal displacement measures resulting from the choice of calibration referent and image rotation in images obtained from video fluoroscopic swallowing studies. In addition, inter- and intrarater reliabilities of hyoid and laryngeal displacement measures were assessed. Finally, measurement variability of displacement measures was used to estimate measurement error for these measures. was noted during the swallow. Selection criteria used for the images included clear visualization of the anterior inferior edge of the hyoid bone (hyoid excursion), the anterior subglottic corner of the air column (laryngeal excursion), the anterior edges of cervical vertebrae C2 C4 (axis of reference and C3 calibration referent), and two custom-made 20-mm calibration referents placed anteriorly and laterally on the participants necks (Fig. 2). Displacement and Length Measurements All measures were made using ImageJ [20]. The application of a vertically aligned C2 C4 axis of reference utilized an image rotation technique previously described [9]. A straight line was drawn between the most anterior inferior points of C2 and C4. Images were then rotated such that the C2 C4 axis line was vertical (Fig. 3). Separate measures were obtained using three different calibration referents: anterior edge of C3 (15 mm), anterior neck referent (20 mm), and lateral neck referent (20 mm). The anterior inferior corner Method Video Fluoroscopic Images Deidentified video fluoroscopic swallow studies (VFSS) from clinical research archives at a tertiary medical center provided the data for the current study. A total of ten image pairs were randomly selected from different patients seen in an outpatient dysphagia clinic. Each image pair originated from a single swallow (Fig. 1). One image of the pairs depicted the subject holding a bolus in the oral cavity before the swallow (preswallow rest position, P1). The other image depicted the point in the swallow reflecting maximal hyolaryngeal excursion (maximum excursion, P2). Maximal excursion was defined as the point where no subsequent anterior or superior hyolaryngeal movement Fig. 2 Video fluoroscopic image with the three calibration referents (lateral neck, anterior neck, and C3) used for hyoid and laryngeal displacement measurement indicated Fig. 1 Video fluoroscopic images showing hyoid and larynx at P1 (left) before swallow and at P2, where hyolaryngeal excursion is maximal (right)

3 Fig. 3 Left A video fluoroscopic image that is unrotated. Right The same image rotated for vertical alignment of the C2 C4 axis of reference of C4 was identified on each image and used as an anchor point. This structure represents a stable (nonmoving) point from which the distance of moving structures (hyoid and larynx) was measured. The anterior inferior corner of the hyoid and the anterior corner of the subglottic air column were used as anatomic landmarks for these structures. Horizontal and vertical displacement measures of the hyoid and larynx were calculated by subtracting values measured on the preswallow image (P1) from those measured on the maximal excursion image (P2). Subsequent to completion of these measures, each image was remeasured without image rotation. In addition to measurements of hyoid and laryngeal displacement, the length of the anterior edge of C3 was measured on each image to validate its use as a comparable calibration referent. The C3 anterior edge was measured first using the anterior neck referent and subsequently the lateral neck referent. To determine the presence of radiographic radial distortion, both external neck referents were measured on each image using the anterior neck referent for calibration. Average measured lengths of the external neck referents were used as an estimate of measurement accuracy. Measurement Reliability To investigate inter-rater reliability, a second rater independently completed all hyoid and laryngeal measurements. To investigate intrarater reliability, both raters remeasured 10% of the measurements 1 week after the initial measurements were obtained. Estimating Measurement Error Measurement error associated with hyoid and laryngeal displacement measures was estimated using the width of the 95% confidence interval for mean displacement values. These estimates were based on measures from rotated images only. Statistical Analysis Statistical analyses were conducted on PASW Statistics 18 (IBM Corporation, Somers, NY) using repeated-measures ANOVA. Displacement was the dependent variable with four levels: hyoid horizontal, hyoid vertical, larynx horizontal, and larynx vertical. The independent variables were referent, image rotation, and rater. t-tests (2-tailed) were used to examine the differences between the measured lengths of (1) the anterior and lateral neck referents and (2) the C3 referent from the arbitrary length of 15 mm. Interand intrarater reliabilities were estimated by intraclass correlation coefficients (ICCs). Results Mean displacement measurements for hyoid and larynx are presented according to calibration referent, image rotation, and rater (Table 1). Greenhouse Geisser adjustment for sphericity violation indicated a single significant interaction between displacement and rotation [F(2.26, 244.1) = 6.69, p = 0.001]. No significant main effect was noted for rater or calibration referent. Planned post-hoc analyses using the least significant difference (LSD) test for the interaction between displacement and rotation revealed a significant difference in horizontal measurements for both the hyoid (p = 0.001) and the larynx (p = 0.05). Displacement measures were greater in the unrotated condition than in the rotated condition (Fig. 4). Vertical measures for both the hyoid and the larynx tended to be smaller in the unrotated condition but these differences were not significant. Mean length of the anterior edge of C3 measured using the anterior and lateral neck referents was mm (SD = 1.56) and mm (SD = 2.35), respectively. Neither of these measured lengths differed significantly from the arbitrarily assigned length of 15 mm (p [ 0.05). Mean lengths of the anterior and lateral neck referents were

4 Table 1 Mean (standard deviation) displacement measurements (rounded to 2 decimal places) according to calibration referent, image rotation, and rater Lateral referent Anterior referent C3 referent Rotated Unrotated Rotated Unrotated Rotated Unrotated Rater 1 Rater 2 Rater 1 Rater 2 Rater 1 Rater 2 Rater 1 Rater 2 Rater 1 Rater 2 Rater 1 Rater 2 Hyoid horizontal 2.44 (5.26) 0.32 (4.31) 2.34 (3.55) 2.82 (3.72) 1.22 (2.79) 0.82 (3.37) 2.65 (2.99) 2.47 (3.25) 2.46 (6.61) 0.66 (6.43) 6.50 (7.72) 2.40 (7.64) Hyoid vertical 5.56 (5.21) 5.10 (5.82) 3.69 (5.47) 5.15 (5.69) 5.01 (5.31) 4.63 (5.17) 3.66 (5.57) 4.34 (5.07) 5.76 (5.51) 5.81 (6.59) 4.66 (5.71) 5.05 (6.36) Larynx horizontal 1.76 (4.53) 0.00 (3.23) 4.23 (4.34) 3.73 (4.12) 0.71 (2.47) 0.35 (2.48) 4.06 (3.31) 3.46 (3.61) 3.70 (11.12) 0.00 (4.55) 6.88 (5.03) 3.72 (3.81) Larynx vertical 9.77 (6.40) (6.71) 9.52 (6.49) (6.34) 9.65 (4.90) 9.26 (5.69) 8.45 (5.53) 9.00 (5.21) (5.38) (5.78) 8.01 (6.47) 9.96 (6.30) Fig. 4 Graph of mean horizontal and vertical displacement measurements for hyoid and larynx in rotated vs. unrotated image conditions Table 2 Estimated measurement error according to the 95% confidence interval of displacement measures based on rotated images Displacement measurement Hyoid horizontal 2.62 Hyoid vertical 2.89 Larynx horizontal 2.48 Larynx vertical mm (SD = 0.15) and 18.3 mm (SD = 0.82), respectively. This difference, though small (an average of 1.64 mm between the two markers), was statistically significant [t(19) = 9.22, p \ 0.001]. Intrarater ICCs for all measures were 0.92 and 0.93 for raters 1 and 2, respectively (p \ 0.001). Inter-rater ICC was 0.77 (p \ 0.001). Using the width of the 95% confidence interval for each measure, measurement error associated with hyoid and laryngeal displacement measures was estimated to range between 2.48 and 3.06 mm (Table 2). Discussion Estimated measurement error (mm) The results of this study indicate that image rotation impacts horizontal hyoid and laryngeal displacement measurements. Conversely, the location of the calibration

5 referent does not significantly impact hyoid or laryngeal displacement measurements. Differences in the length of the anterior versus the lateral neck referent demonstrated the radial image distortion effect (i.e., the anterior referent closer to the image edge was slightly larger). However, this effect is of small magnitude and hyoid and laryngeal displacement measures were not affected by the location of the referent used. Thus, it is reasonable to conclude that measurements made using different referents are comparable for the measurements examined in this study. Measurements of the anterior edge of the C3 referent also supported the assignment of 15 mm as the length of this structure [11, 12]. Thus, an anterior length at 15 mm for C3 may be an acceptable calibration referent when no external referents are available. Although the use of image rotation to align the C2 C4 vertically has been commonly used to control for differences in head positions, no study has reported on the effect of the image rotation technique on hyoid and laryngeal displacement measures. Our results suggest that only horizontal measures are affected significantly. Vertical measures (e.g., elevation of each structure) are not impacted by image rotation. Various approaches have been discussed regarding the axis of reference for hyoid and laryngeal displacement measurements. Some researchers have used axes of reference other than C2 C4 [10, 18, 19]. Other researchers chose to measure excursion distance instead of separate horizontal and vertical displacements and excluded the use of an axis of reference altogether [21]. Our results emphasize that studies of hyoid and larynx displacement during swallowing cannot be compared easily if methodological approaches differ. This position supports the argument put forth by Zu et al. [19] that even slight differences in methodologies may affect resulting measurements. Strong reliability within and between raters was confirmed by the ICC measures. In this study, intrarater reliability is comparable to that of prior studies of hyoid and laryngeal kinematics in healthy participants [1, 9, 11, 15]. Inter-rater reliability is lower than those reported in prior studies in healthy participants but similar to those of studies of patients with dysphagia [5, 12]. The strong agreement between raters in this study attests to the robustness of the image analysis approach described for measuring hyoid and laryngeal displacement during swallowing. Only small differences (margin of error) were observed when measuring objects of known lengths (i.e., anterior and lateral neck referents). For example, the difference between the measured lengths and the actual lengths was 0.08 mm for the anterior neck referent compared with 1.72 mm for the lateral neck referent. These results compare well with measurement accuracy reported in other studies of radiographic image measurement, such as in conventional and digital radiography (1.5 and 1.0 mm, respectively) [22], indirect and direct computed radiography (1.4 and 2.6 mm, respectively) [23], and CT imaging of the skull base (1 mm) [24]. In the present study, measurement error associated with hyoid and laryngeal displacement measures was estimated to be between 2.48 and 3.06 mm. Ludlow et al. [25] speculated that a 2 3-mm change in laryngeal elevation seen in their participants was due to measurement variability. However, these investigators did not provide data clarifying their assumption of measurement error. Investigators in future studies reporting and interpreting measurements of hyoid and laryngeal displacement should give consideration to an estimated margin of error when interpreting results. Measurements in the current study were obtained from video fluoroscopic images of patients with dysphagia. In healthy participants, a greater range of hyoid and laryngeal displacement during swallowing may contribute to a greater degree of variability [26]. Generalization of the current results to studies on healthy subjects should therefore be done with caution. In addition, the present results may not apply directly to other aspects of swallowing kinematics such as measurement of the pharyngeal constriction ratio [27], which involves an area measurement rather than a distance measurement. This study demonstrated that the use of different calibration referents did not contribute significantly to variability in measures of hyoid and laryngeal displacement but that image rotation affected horizontal measures for these structures. Additionally, we provided evidence supporting the use of C3 at 15 mm as an anatomical calibration referent. Potential measurement error in hyoid and larynx displacement measures was quantified and serves as a guide to investigators for better interpretation of hyoid and laryngeal displacement measures in future studies. References 1. Logemann J, Pauloski B, Rademaker A, Kahrilas P. Oropharyngeal swallow in younger and older women: videofluoroscopic analysis. J Speech Lang Hear Res. 2002;45: Paik N, Kim S, Lee H, Jeon J, Lim J, Han T. Movement of the hyoid bone and the epiglottis during swallowing in patients with dysphagia from different etiologies. J Electromyogr Kinesiol. 2008;18: Ravi B, Rampersaud R. Clinical magnification error in lateral spinal digital radiographs. Spine (Phila Pa 1976). 2008;33:E Cerveri P, Forlani C, Borghese N, Ferrigno G. Distortion correction for X-ray image intensifiers: local unwarping polynomials and RBF neural networks. Med Phys. 2002;29: Kim Y, McCullough GH. Maximal hyoid excursion in poststroke patients. Dysphagia. 2010;25:20 5.

6 6. Logemann J, Kahrilas P, Begelman J, Dodds W, Pauloski B. Interactive computer program for biomechanical analysis of videoradiographic studies of swallowing. AJR Am J Roentgenol. 1989;153: Crary M, Butler M, Baldwin B. Objective distance measurements from videofluorographic swallow studies using computer interactive analysis: technical note. Dysphagia. 1994;9: Perlman A, VanDaele D, Otterbacher M. Quantitative assessment of hyoid bone displacement from video images during swallowing. J Speech Hear Res. 1995;38: Logemann J, Pauloski B, Rademaker A, Colangelo L, Kahrilas P, Smith C. Temporal and biomechanical characteristics of oropharyngeal swallow in younger and older men. J Speech Lang Hear Res. 2000;43: Ishida R, Palmer J, Hiiemae K. Hyoid motion during swallowing: factors affecting forward and upward displacement. Dysphagia. 2002;17: Kim Y, McCullough G. Maximum hyoid displacement in normal swallowing. Dysphagia. 2008;23: Pauloski B, Logemann J, Fox J, Colangelo L. Biomechanical analysis of the pharyngeal swallow in postsurgical patients with anterior tongue and floor of mouth resection and distal flap reconstruction. J Speech Hear Res. 1995;38: Kim S, Han T, Kwon T. Kinematic analysis of hyolaryngeal complex movement in patients with dysphagia development after pneumonectomy. Thorac Cardiovasc Surg. 2010;58: Ekberg O. The normal movements of the hyoid bone during swallow. Invest Radiol. 1986;21: Kendall K, Leonard R, McKenzie S. Accommodation to changes in bolus viscosity in normal deglutition: a videofluoroscopic study. Ann Otol Rhinol Laryngol. 2001;110: Wang TG, Chang YC, Chen WS, Lin PH, Hsiao TY. Reduction in hyoid bone forward movement in irradiated nasopharyngeal carcinoma patients with dysphagia. Arch Phys Med Rehabil. 2010;91: Matsuo K, Palmer JB. Kinematic linkage of the tongue, jaw, and hyoid during eating and speech. Arch Oral Biol. 2010;55: Kellen PM, Becker DL, Reinhardt JM, Van Daele DJ. Computerassisted assessment of hyoid bone motion from videofluoroscopic swallow studies. Dysphagia. 2010;25: Zu Y, Yang Z, Perlman AL. Hyoid displacement in post-treatment cancer patients: preliminary findings. J Speech Lang Hear Res. 2011;54(3): Rasband WS. ImageJ, ed. Bethesda, MD: National Institutes of Health, Leonard R, Kendall K. Dysphagia assessment and treatment planning: a team approach. 2nd ed. San Diego: Plural Publishing; Swennen GR, Grimaldi H, Berten JL, Kramer FJ, Dempf R, Schwestka-Polly R, Hausamen JE. Reliability and validity of a modified lateral cephalometric analysis for evaluation of craniofacial morphology and growth in patients with clefts. J Craniofac Surg. 2004;15: (discussion 413 4). 23. Fowler JR, Ilyas AM. The accuracy of digital radiography in orthopaedic applications. Clin Orthop Relat Res. 2011;469(6): Yang HA, Yang Y, Wang HW, Meng QL, Ren XH, Liu YG. A comparative study of digital and anatomical techniques in skull base measurement. J Int Med Res. 2010;38: Ludlow C, Humbert I, Saxon K, Poletto C, Sonies B, Crujido L. Effects of surface electrical stimulation both at rest and during swallowing in chronic pharyngeal. Dysphagia. 2007;22: Lof GL, Robbins J. Test-retest variability in normal swallowing. Dysphagia. 1990;4: Leonard R, Rees CJ, Belafsky P, Allen J. Fluoroscopic surrogate for pharyngeal strength: the pharyngeal constriction ratio (PCR). Dysphagia. 2011;26:13 7. Isaac Sia BSc Pamela Carvajal MA Giselle D. Carnaby-Mann PhD, MPH, BSc, PGrad Dip (HSC), SLP-CCC Michael A. Crary PhD, FASHA