Three-Dimensional Endoscopic Images of Vocal Fold Paralysis by Computed Tomography

Objective  To describe characteristics of 3-dimensional (3-D) computed tomographic (CT) endoscopic images of the larynx in unilateral vocal fold paralysis and the changes of the paralyzed vocal fold after phonosurgery as indicated by 3-D CT endoscopy.

Setting  A university medical center.

Patients  Twelve consecutive patients with unilateral vocal fold paralysis who underwent helical CT examination. Five of them underwent the CT examination before and after phonosurgical treatment.

Interventions  Three patients underwent arytenoid adduction, and 2 underwent type 1 thyroplasty.

Main Outcome Measures  "Sagging," defined as caudal displacement of the vocal fold; and "thinning," defined as a decrease in the vertical thickness of the vocal fold and expansion of the ventricle on the affected side, were evaluated.

Results  Excessive motion artifacts in one patient prevented detailed description of his 3-D images. Sagging and thinning of the vocal fold and expansion of the ventricle on the affected side were noted on 6, 11, and 8 occasions, respectively. Adduction or augmentation of the paralyzed vocal fold after phonosurgery was observed in 3-D CT endoscopic images when displayed with bony densities.

Conclusions  The use of 3-D CT endoscopy enables description of 3-D characteristics of unilateral vocal fold paralysis and supplements stroboscopic findings. Furthermore, CT endoscopic images, when simultaneously displayed with bony densities, may help in evaluating the effects of phonosurgical treatment of the paralyzed vocal fold.

UNILATERAL vocal fold paralysis (UVFP), which causes breathy dysphonia or mild aspiration, is an important clinical entity in otolaryngology. Surgical treatment for UVFP includes lateral compression of the paralyzed vocal fold,¹ adduction of the arytenoid cartilage,² and intracordal injection of various biocompatible materials such as collagen^3,4 and autogenetic fat.^5,6 Reinnervation procedures, such as anastomosis of the ansa cervicalis to the recurrent laryngeal nerve⁷ and ansa cervicalis neuromuscular pedicle implantation into the thyroarytenoid and lateral cricoarytenoid muscles,⁸ have also been reported. The paralyzed vocal fold may be fixed at the midline, paramedian, or intermediate position. In addition to the position of the paralyzed vocal fold, atrophy of the fold and positional differences in the superior-inferior (vertical) direction between bilateral vocal folds during phonation should be corrected by surgical treatment.

Videostroboscopy combined with an oblique-view rigid endoscope or a flexible fiberscope has been most often used to assess vocal fold vibration before and after surgical treatment.^4,6,9,10 In most patients with UVFP, such examination can visualize position of the paralyzed vocal fold, degree of atrophy, and presence or absence of the glottal gap during phonation. However, distortion of images derived from the optical system of the endoscope used and positional differences in the vertical direction between bilateral vocal folds may prevent accurate assessment of the pathological changes in the paralyzed vocal fold. For example, the vocal fold on the paralyzed side may appear to be longer than that on the healthy side because the paralyzed vocal fold is often situated in a higher position during phonation¹¹ and is closer to the tip of the endoscope. In addition, overadduction of the ventricular folds and anterior-posterior shortening of the glottis may occur during phonation. In such patients, the vocal fold cannot be visualized from the oral side and videostroboscopy does not provide useful information regarding the vocal fold during phonation.

The recent development of a helical-scanning high-resolution computed tomographic (CT) and 3-dimensional (3-D) reconstruction technique has allowed visualization of the interiors of great vessels and luminal organs.¹² Yumoto et al¹³ applied this 3-D CT endoscopic technique to assess normal laryngeal structures and pathological changes as exemplified by laryngeal cancer, UVFP, and atrophy of the vocal fold. They produced laryngeal images viewed from the oral and tracheal sides as well as 2 vertically split hemilaryngeal images. With the exception of those from the oral view, such reconstructed images are unobtainable through indirect mirror examination and endoscopic observation. They might be able to 3 dimensionally depict characteristics of UVFP that are not detected with conventional laryngeal observations. The present study describes characteristics of 3-D CT endoscopic images of the paralyzed larynx and the changes of the paralyzed vocal fold after phonosurgery as indicated by 3-D CT endoscopy. Since Yumoto et al¹³ reported the usefulness of axial and coronal images produced by the multiplanar reconstruction (MPR) method, and since these sectional images show fine resolution between bone and soft tissue and within the soft tissue, we evaluated 3-D CT endoscopic images combined with axial and coronal images.

The subjects included 7 men and 5 women with UVFP who underwent helical CT examination between April 16, 1996, and November 6, 1997, to produce 3-D endoscopic images. Their ages ranged between 21 and 70 years (mean, 47 years). The clinical profiles of the subjects are shown in Table 1. Five of the 12 subjects underwent CT examination before and after phonosurgical intervention. All except subjects 2, 3, and 6 received videostroboscopic and aerodynamic examinations during sustained phonation within 1 week of the CT scan. Observation of the larynx was performed with videostroboscopy on each occasion. Aerodynamic studies included determination of mean airflow rate, AC/DC ratio (an index of vocal efficiency),¹⁴ and maximum phonation time, which were measured during sustained phonation of the vowel /a/ at a comfortable pitch and loudness (Phonation Analyzer PA500; Minato Co, Tokyo, Japan).

The scanning technique and production of 3-D endoscopic and coronal MPR images are described herein only briefly as the details were reported previously.¹³ Subjects lay in the supine position with a pillow under their neck to avoid overlapping of the mandible above the larynx. All scans were performed with a helical-scanning high-speed x-ray CT scanner (X-Vigor; Toshiba, Tokyo) using 150 mA, 135 kV, 1-mm collimation, and 1-mm/s table speed. Each patient was asked to hold his or her breath for approximately 30 seconds following near-maximal inspiration, and was scanned from the protrusion of the superior thyroid notch to approximately 30 mm below this point so that the scanning range included the ventricular fold, ventricle, vocal fold, and subglottic region.

Axial images were generated at 1-mm intervals by overlapping each adjacent image by 0.5 mm and then sent via a local area network to a workstation (X-Tension; Toshiba) for further processing. A volume-rendering technique was used with 3-D endoscopic mode that classifies voxels (unit of volumetric data) with 2 numbers that specify the upper and lower limits (thresholds) to display a surface of specified anatomical parts in the 3-D image. In our study, −100 Hounsfield units was the upper threshold and −600 Hounsfield units was the lower threshold. Voxels outside the range defined by the 2 thresholds were eliminated. Views were displayed from the tracheal and oral sides as well as 2 vertically split hemilaryngeal images. In subjects 8 to 12, who underwent phonosurgical intervention, voxels between 150 and 1500 Hounsfield units were also extracted and simultaneously displayed with the endoscopic images. These thresholds were set to depict the bony structures. Since ossification of the laryngeal cartilages varies among the sites of interest within a subject as well as among subjects, the images could show the outline of the spatial relationship between the vocal fold and laryngeal framework including the arytenoid cartilage. In subject 5, an additional view was obtained to demonstrate pathological changes. Coronal MPR images (3 mm thick) perpendicular to the glottic axis were systematically built in all subjects.

The presence or absence of "stairstep" motion artifacts and glottic aperature during each scan were evaluated on CT endoscopic images. We evaluated the presence or absence of "sagging" and "thinning" of the paralyzed vocal fold, and expansion of the ventricle on the affected side. Sagging was defined as caudal displacement of the membranous vocal fold on the affected side. Thinning was defined as a decrease in the vertical thickness of the paralyzed fold by comparison with that of the healthy side. Figure 1, A, shows CT endoscopic images of subject 10 (CT scan 12) exemplifying sagging and thinning of the fold and expansion of the ventricle on the affected side.

Computed tomographic images obtained from subjects 8 through 12 before and after phonosurgical treatment were compared to describe the changes of the paralyzed vocal fold after phonosurgical intervention.

Table 2 summarizes CT endoscopic (presence or absence of stairstep motion artifacts, glottic aperature during scanning, sagging and thinning of the fold, and expansion of the ventricle on the affected side) and videostroboscopic findings. Of the 17 CT examinations performed, stairstep motion artifacts occurred on 6 occasions, although they did not prevent production of 3-D endoscopic images. Figure 1 and Figure 2 exemplify CT endoscopic images without (subject 10) and with (subject 12) motion artifacts, respectively. Excessive motion artifacts seen in subject 8 (CT scan 9) prevented detailed description of his 3-D images. On 9 occasions, the glottis was closed during scanning so that hemilaryngeal images could not show the detailed structure of the laryngeal lumen. However, even when the glottis was closed during scanning, the lower surface of the vocal fold was seen so that the thickness of the lower portion of the vocal fold could be evaluated. Thinning of the paralyzed fold was identified on 11 occasions. Figure 1, B, shows CT endoscopic images of subject 10 (CT scan 13), whose glottis during scanning was closed. However, thinning of the paralyzed fold was noted in the hemilaryngeal images. Sagging of the vocal fold and expansion of the ventricle on the affected side were noted on 6 and 8 occasions, respectively.

On 4 of the 14 stroboscopic examinations performed, overadduction of the ventricular fold, anterior-posterior shortening of the glottis, or both precluded assessment of the paralyzed fold. Atrophy of the paralyzed fold was noted on 4 of 10 occasions evaluated. Figure 3, A, shows conventional endoscopic images of subject 8 during inspiration (left) and phonation (right) before surgery. Although the left vocal fold did not show atrophic changes endoscopically, CT endoscopic hemilaryngeal images revealed decreased thickness of the left vocal fold (Figure 3, B).

Table 3 presents measurements of mean airflow rate, AC/DC ratio, and maximum phonation time in subjects 8 through 12. All aerodynamic measurements except maximum phonation time in subject 9 were improved after surgery. The aerodynamic measurements and CT endoscopic findings were compared preoperatively and postoperatively in subjects 9 through 12. Subject 8 was excluded because of excessive motion artifacts. In subject 12, thickness of the vocal fold, as assessed by CT endoscopic images, seemed to have become symmetrical after surgery (Figure 2), while in the other 3, it remained decreased on the paralyzed side even after surgery. In subject 12, sagging of the paralyzed fold also disappeared after surgery.

Computed tomographic endoscopic images viewed from the tracheal side in subject 10 revealed that the arytenoid cartilage on the paralyzed side was rotated to adduct the vocal fold after surgery (Figure 1). In subject 12, axial images showed rotation of the arytenoid cartilage (Figure 2, C), although this was not prominent in CT endoscopic images. Figure 4 and Figure 5 show preoperative and postoperative CT endoscopic and coronal MPR images for subjects 9 and 11, respectively. The vocal fold in these subjects was not visualized endoscopically because of overadduction of the ventricular fold, anterior-posterior shortening of the glottis, or both. A CT endoscopic image viewed from the tracheal side in subject 9 before surgery indicated that the left vocal fold was thinner than the right. He underwent type 1 thyroplasty on the left side. Postoperative CT endoscopic images showed that a silicone block augmented the left vocal fold (Figure 4, A). A coronal MPR image confirmed that the silicone block was inserted at the level of the vocal fold (Figure 4, B). In subject 11 who did not show improvement of dysphonia after type 1 thyroplasty and arytenoid adduction, the coronal MPR images indicated that a block of silicone was situated slightly more cranially than the vocal fold (Figure 5, B). Type 1 thyroplasty was performed again to replace the silicone block to a more caudal position. The coronal MPR images taken after the second type 1 thyroplasty (Figure 5, B) showed that the replaced silicone block was properly located at the level of the vocal fold. Computed tomographic endoscopic images viewed from the tracheal side indicated that the lower surface of the right vocal fold was augmented after replacement of the silicone block (Figure 5, A).

Although videostroboscopy has been widely used to evaluate pathological changes in the paralyzed larynx, especially during phonation, some patients do not allow observation of the larynx through an endoscope because of overadduction of the ventricular fold, anterior-posterior shortening of the glottis, or both. In the present series, 3 subjects (4 of 14 occasions performed) showed such strangulation of the laryngeal inlet. In addition, intolerance to insertion of the rigid endoscope or fiberscope may occur, precluding observation of the larynx. In the present series, 1 (subject 3) of the 12 subjects could not tolerate endoscopic examination. Thus, videostroboscopy could not provide useful diagnostic information in these 4 subjects (5 occasions). Aerodynamic measurements may help in estimating the size of the gap between the vocal folds during phonation in these subjects.

The purpose of videostroboscopy is to assess glottic closure, mucosal wave, vertical level of vocal fold approximation, and atrophy and position of the paralyzed vocal fold.^9,15 Although videostroboscopy is the only method to assess the glottis and mucosal wave of the vocal fold during phonation, it has some disadvantages. For example, which side was situated higher could not be assessed because the endoscopes were not binocular and did not provide depth information. In addition, a previous report¹⁶ indicated that assessment of paralyzed vocal fold position based on endoscopic observation was inconsistent among different judges.

Radiological examinations include conventional laryngeal tomography,^11,17 laryngography,¹⁸ and CT.^13,18,19 The former 2 methods are able to detect the presence of UVFP but do not depict detailed morphologic changes of the paralyzed vocal fold. These methods were used to diagnose UVFP by detecting elevation of the vocal fold during phonation and expansion of the ventricle on the affected side for patients in whom conventional laryngoscopy was not possible for various reasons. Agha¹⁸ reported that laryngography revealed distinct features of UVFP compared with conventional CT. Zeiberg et al¹⁹ reported that the ventricle could not be identified in any of their 3-D CT images obtained using 5-mm sections. Computed tomographic imaging of the larynx requires the use of thinner sections to resolve small anatomical structures. Yumoto et al¹³ first succeeded in producing 3-D CT endoscopic images of the laryngeal structures in detail. They suggested that 3-D CT endoscopic images combined with cross-sectional (coronal and axial) images could offer diagnostic information in patients with UVFP.

We displayed the laryngeal structures viewed from the oral and tracheal sides. Furthermore, the hemilaryngeal images viewed from the other side provided distinct depictions of the larynx. With the exception of those from the oral side, these views cannot be obtained by any other method used for observation of the larynx. Tracheal views were useful when observation of the larynx through the endoscope was not possible because of strangulation of the laryngeal inlet or intolerance to insertion of the endoscope. In addition, the position of the paralyzed vocal fold was assessed in the tracheal view more accurately than in the oral view because the whole length of the vocal fold, which is composed of the membranous and cartilaginous parts, could be visualized. As shown in Figure 2, B, and Figure 3, B, the cartilaginous part was not always included in the images viewed from the oral side. This situation also occurred frequently on conventional endoscopic observation.

The 3-D CT hemilaryngeal images enabled evaluation of sagging of the vocal fold and expansion of the ventricle on the affected side. These features were thought to indicate loss of or a decrease in tension of the intrinsic laryngeal muscles. Loss of pull by the posterior cricoarytenoid muscle could result in anterior rotation of the arytenoid cartilage and caudal displacement of the vocal process. Furthermore, elastic character of the conus elasticus may facilitate caudal displacement of the membranous vocal fold. Also, thinning of the paralyzed fold could be evaluated in the hemilaryngeal images by comparison with the healthy side. This evaluation of vertical vocal fold thickness was possible even when the glottis was closed during scanning, as exemplified in Figure 1, B. Five subjects (subjects 3-6 and 12 after surgery) did not show thinning of the paralyzed fold. Each of these 5 subjects underwent phonosurgical treatment: subjects 3 and 12 underwent arytenoid adduction; subjects 4 and 6 had severed recurrent laryngeal nerves anastomosed; and in subject 5, silicone was injected into the paralyzed fold. The absence of thinning of the paralyzed fold in these subjects may have been due to the phonosurgical treatments. On the other hand, only 1 (subject 12) of the 4 subjects who underwent the CT examination before and after phonosurgery showed postoperative symmetry in the vertical thickness of the vocal fold. Further study combined with electromyographic investigation is required to determine whether thinning was a result of atrophy of the intrinsic laryngeal muscles or loss of or a decrease in tension of the muscles.

Figure 6 shows CT endoscopic images from subject 5, who came to our clinic complaining of a 1-year history of dysphonia. His voice had been satisfactory for 4 years since he had undergone room temperature–vulcanizing silicone injection to the paralyzed left vocal fold at another institute 5 years previously. Voxels between 150 and 1500 Hounsfield units were simultaneously displayed with the CT endoscopic images. Injected silicone was seen as a mass of particles dispersed in the soft tissues of the left vocal fold, and migration of silicone in the vocal fold was considered to have occurred. Tsuzuki et al²⁰ histologically examined the human larynx 12 years after room temperature–vulcanizing silicone injection into the paralyzed fold. They reported that silicone particles were encapsulated by thin fibrous tissue and that many particles were present in the vocalis muscle. Figure 6 also shows that the left vocal fold was swollen compared with that of the healthy side. This change was assumed to have become prominent 4 years after silicone injection. Although there have been no reports of granuloma formation after silicone injection into the vocal fold, the late effects of silicone injection in subject 5 suggest that granuloma formation is a possible late adverse effect, similar to Teflon.²¹ Thus, CT endoscopy can help in estimating the varying causes of dysphonia in patients with UVFP.

One disadvantage of CT endoscopy is the occasional occurrence of stairstep motion artifacts. In fact, excessive motion artifacts precluded evaluation of vertical thickness of the vocal fold in subject 8 (CT scan 9). Since subjects are required to hold their breath for approximately 30 seconds during scanning, those in whom pulmonary functions are poor cannot be examined. Another disadvantage is that, at present, helical CT scanners are not fast enough to allow scanning of the larynx during phonation. Quantitative assessment of the features observed in the paralyzed larynx is an issue to be raised. However, at present, it is not possible to determine even the glottic axis in 3-D endoscopic images as a standard to which we could objectively express the position of the paralyzed vocal fold. Further studies are required to optimize the CT endoscopic technique for its application to UVFP and to quantitatively assess the 3-D endoscopic features of the paralyzed larynx.

Although the number of subjects in the present study was relatively small, the use of 3-D CT endoscopy combined with examination of coronal and axial images enables description of 3-D characteristics of the paralyzed larynx and supplements videostroboscopic findings. Furthermore, CT endoscopic images simultaneously displayed with bony density may help in evaluating the effects of phonosurgical treatment on the paralyzed vocal fold.

Accepted for publication March 11, 1999.

Reprints: Eiji Yumoto, MD, Department of Otolaryngology–Head and Neck Surgery, Kumamoto University School of Medicine, 1-1-1 Honjo, Kumamoto 860-8556, Japan (e-mail: yumoto@gpo.kumamoto-u.ac.jp).