OR I G I N A L A R T I C L E

Intraoperative image guidance in transoral robotic surgery:A pilot study

Andrew K. Ma, MD1 | Michael Daly, PhD2 | Jimmy Qiu, MASc2 |

Harley H. L. Chan, PhD2 | David P. Goldstein, MD, MSc, FRCSC1 |

Jonathan C. Irish, MD, MSc, FRCSC1,2 | John R. de Almeida, MD, MSc, FRCSC1

1Department of Otolaryngology – Head and

Neck Surgery/Surgical Oncology, University

of Toronto, Toronto, Ontario, Canada

2Guided Therapeutics (GTx) Program,

Techna Institute, University Health Network,

Toronto, Ontario, Canada

Correspondence

John R. de Almeida, Princess Margaret

Cancer Centre, 610 University Avenue,

3-955, Toronto, Ontario, Canada M5G

2M9.

Email: john.delameida@uhn.ca

Funding information

This project was made possible through

funding provided by the Harry Barberian

Scholarship, Department of Otolaryngology,

University of Toronto, Toronto, Ontario,

Canada.

Abstract

Background: Intraoperative image guidance during transoral robotic surgery (TORS) is hampered

by imaging-friendly instrumentation and intraoperative positioning. The purpose of this study was

to develop and validate an accurate image-guidance system for TORS.

Methods: A custom radiolucent mouth retractor was fabricated from biocompatible material

(Med-610; Stratasys, Minneapolis, MN). Teflon beads were placed in the oropharynx and carotid

arteries of 3 cadavers. CT scans were obtained in the preoperative and intraoperative positions.

Displacement of targets between preoperative and intraoperative scans was measured. Surgical

navigation was based on the open-source Image-Guided Surgery Toolkit. Target registration error

(TRE) was determined by measuring the distance between the tracker and bead registered to pre-

operative versus intraoperative scans.

Results: The inferior oropharyngeal targets demonstrated the greatest displacement between

positions. A significant reduction in TRE was observed when registering the tracker to the intrao-

perative compared to the preoperative scan.

Conclusion: This study describes an accurate intraoperative image-guidance system for TORS.

K E YWORD S

cadaver study, head and neck cancer, image guidance, intraoperative CT, transoral robotic surgery

(TORS)

1 | INTRODUCTION

Transoral robotic surgery (TORS) has several applications in the surgical

management of oropharyngeal cancers and in selected laryngeal, hypo-

pharyngeal, parapharyngeal, and nasopharyngeal lesions.1 TORS obvi-

ates the need for transmandibular approaches to provide access for

tumor resection. Several studies have demonstrated the efficacy of

TORS in treatment of early-stage oropharyngeal cancers (T1-T2), with

locoregional control rates similar to that of traditional chemoradiation

while avoiding its late toxicities.2–4 In addition to functional and onco-

logic benefits of this approach, TORS provides a high resolution magni-

fied endoscopic view, the use of angled endoscopes to navigate

around corners, the ability to maintain maximal degrees of freedom

within a narrow space, and the elimination of surgeon hand tremor.3,5

However, the transoral robotic approach is limited in several ways.

First, proximity of tumors to critical vascular structures, such as the

internal carotid artery in the parapharyngeal space, may compromise

adequate oncologic resection. Furthermore, the lymphoid appearance

of the tongue base can obscure proper identification of deep, submu-

cosal tumors that may ordinarily be identified by manual palpation.

TORS lacks the haptic feedback that a surgeon relies on in standard

approaches for proper surgical resection. These limitations may result

in inadequate resections or possibly even catastrophic bleeding.

Image guidance, typically used for endoscopic sinus surgery or

skull base surgery, can provide improved surgeon comfort with vascular

anatomy and real-time imaging corroboration of endoscopic visualiza-

tion with direct anatomic localization on triplanar imaging views with

the use of a tracking device. Applying these systems is relatively

1976 | VC 2017Wiley Periodicals, Inc. wileyonlinelibrary.com/journal/hed Head Neck. 2017;39:1976–1983.

Received: 9 August 2016 | Revised: 25 January 2017 | Accepted: 28 February 2017

DOI: 10.1002/hed.24805

straightforward and predictable when registering to static bony land-

marks. However, the jaw, lips, and tongue must be retracted in a trans-

oral approach in order to access the tumor. Deformation of soft tissue

anatomy with intraoperative jaw positioning compromises accuracy of

registration to the preoperative imaging study. Because of these posi-

tional changes, an image-guidance system and tracking device cannot

be registered to preoperative scans that are obtained with the mouth

closed. This study validates the proof of principle in the developed

instrumentation, in addition to the accuracy of the image-guidance

system for use with TORS.

2 | MATERIALS AND METHODS

2.1 | ETHICS APPROVAL

Research ethics board approval was obtained for cadaveric studies.

2.2 | Cadaveric specimens and internal targets

Three cadaveric head specimens were obtained through the University

of Toronto Surgical Skills Centre (Mount Sinai Hospital, Toronto, Can-

ada). These specimens were preserved in a fresh-frozen preparation to

most accurately represent natural soft tissue movement in a live

patient; as formalin preserved specimens demonstrate significant soft

tissue rigidity.

Specimens were selected based on the following criteria: (1)

adequate mouth opening to accommodate the radiolucent mouth

retractor; (2) no evidence of oral cavity or oropharyngeal masses; and

(3) no evidence of oral cavity or oropharyngeal instrumentation or sur-

gery. In some specimens, dental extraction was performed on the ante-

rior maxillary and mandibular dentition to obtain more interincisal

distance to accommodate the radiolucent mouth retractor. Figures 1

and 2 demonstrate positioning of the cadaveric specimen.

Four-millimeter diameter Teflon beads were selected as targets for

the study (see Figure 3). This material demonstrates bright enhance-

ment distinct from soft tissue on noncontrast CT and presents minimal

streak artifact. The Teflon beads were drilled and mounted onto

0.01600 diameter stainless steel posts. The stainless steel posts were

tested through the CT scanner and also demonstrated minimal streak

artifacts. The targets were then tacked onto a variety of oropharyngeal

subsites and demonstrated stable seating on repeated mouth opening

and closing.

Oropharyngeal subsites were as follows: (a) uvula (31); (b) superior

tonsillar fossa (32); (c) inferior tonsillar fossa (32); (d) posterolateral

oropharyngeal wall (32); (e) base of tongue (32); and (f) epiglottis (31).

Three targets were also placed into each internal carotid artery. A

vertical incision was made on each side of the neck over the trajectory

of the common carotid artery. This was dissected until the common

carotid artery, carotid body, and internal and external branches were

exposed. The internal carotid artery was identified. Three Teflon beads

were placed into each carotid system, from the level of the hyoid infe-

riorly and up to the internal carotid artery in the parapharyngeal space

deep to the mandible.

2.3 | Image-guided surgery system

The image-guidance system includes a mobile C-arm for intraoperative

cone-beam CT, real-time optical tracking, and custom 3D visualization

software (see Figure 4). The flat-panel cone-beam CT prototype system

was developed in collaboration with Siemens Healthcare (Erlangen,

Germany) to enable 3D volumetric imaging in a single half-rotation

(approximately 1808) without translation of the patient. A recent pro-

spective clinical study in head and neck surgery demonstrates that

intraoperative cone-beam CT provides high-quality 3D imaging during

bone ablation and reconstruction tasks with acceptable workflow inter-

ruptions.6 For the current study, intraoperative acquisition consisted of

200 X-ray projections at 100 kVp obtained in approximately 60 sec-

onds, resulting in volumetric images encompassing approximately 20 3

20 3 15 cm3 with isotropic voxels (0.8 mm3). Radiation doses for

cone-beam CT-guided head and neck surgery have been reported pre-

viously at levels sufficiently low (ie, approximately one-fifth of a typical

2-5 mSv diagnostic head CT) to permit repeat intraoperative imaging.7

Navigation of surgical instruments (eg, pointers, endoscopes, and drills)

is performed using an infrared stereoscopic camera (Polaris Spectra;

NDI, Waterloo, Canada) and corresponding retroreflective spheres. The

in-house navigation software (“GTx-Eyes”), based on the open-source

image-guided surgery toolkit8,9 provides a variety of 3D visualization

capabilities, including standard triplanar views (axial/sagittal/coronal),

semitransparent surface renderings, 3D point localization, and image

fusion.

FIGURE 1 Cadaveric specimen in the preoperative (mouth closed)position [Color figure can be viewed at wileyonlinelibrary.com]

MA ET AL. | 1977

2.4 | Radiolucent mouth retractor

A radiolucent mouth retractor was fabricated through the Princess

Margaret Hospital Machine Shop (Toronto, Canada). The design was

based on the existing Feyh-Kastenbauer retractor. The retractor was

built using a proprietary radiolucent material, Med-610 (Stratasys, Eden

Prairie, MN), which is biocompatible and autoclavable.

2.5 | Imaging procedure

Surface fiducials were used to register specimens to CT scans of the

cadaveric head specimens obtained in either the mouth closed (preop-

erative) or mouth open (intraoperative) positions. Surface fiducials

were placed in the preauricular, zygomatic, infraorbital, and nasion

positions (see Figure 1). Alignment errors of <1 mm were achieved.

Cone-beam CT scans (noncontrast) of the cadaveric head speci-

mens with oropharyngeal targets were obtained in 2 positions: (1)

mouth closed (see Figure 1); and (2) mouth open with radiolucent

retractor in place (see Figure 2). The first position represents preopera-

tive CT scanning, whereas the second position represents intraopera-

tive positioning for TORS with the mouth opened and the retractor in

position. Surface fiducials were placed in the preauricular, zygoma,

infraorbital, and nasion positions, and used for image registration, as

described below. The experimental setup is shown in Figure 4.

2.6 | Anatomic tissue displacement

The amount of anatomic tissue displacement was determined by

comparing the mouth closed versus the mouth open position CT

scans, after obtaining correct alignment through paired-point

registration of surface fiducials. The centroid of the oropharyngeal

and internal carotid artery targets were marked in the axial, coronal,

and sagittal CT cuts using GTx-Eyes software. The center position

was confirmed visually with magnified cross-sectional imaging.

In this fashion, the targets were marked in order for both the

preoperative and intraoperative positions. Using MATLAB (R2014;

MathWorks, Natick, MA), the 3D displacement of each target was

calculated. Selection, labeling, and verification of all targets were

done by one author (A.M.).

2.7 | Tracking system accuracy

The target registration error (TRE) at each of the oropharyngeal targets

was determined by using a bayonet tracking instrument registered to

the preoperative and intraoperative CT scans via surface fiducials. The

instrument was then used to physically contact a Teflon target in

the intraoperative position under direct visualization while recording

the position of the instrument. Each target was contacted 5 times by

the tracking instrument and recorded. The 3D distance between the

recorded position of the tracked instrument at each specific target

and the corresponding position of the target according to CT was

determined. In this fashion, the 3D distance between the tracker and

targets was used to determine TRE when registered separately to the

preoperative and intraoperative positions.

FIGURE 2 Cadaveric specimen in the intraoperative (mouthretracted) position [Color figure can be viewed atwileyonlinelibrary.com]

FIGURE 3 Teflon beads inset as oropharyngeal targets [Colorfigure can be viewed at wileyonlinelibrary.com]

1978 | MA ET AL.

FIGURE 4 Experimental set up with C-arm cone-beam CT scanner and cadaveric specimen with radiolucent mouth retractor in position[Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 5 Comparison of a metal Feyh-Kastenbauer retractor versus a modified radiolucent retractor. A, Positioning of the metal Feyh-Kastenbauer retractor in the phantom skull model. B, CT scan of the metal Feyh-Kastenbauer retractor demonstrating considerable streakartifact. C, Positioning of the radiolucent retractor in the cadaveric model. D, CT scan of the radiolucent retractor demonstrates no artifacts[Color figure can be viewed at wileyonlinelibrary.com]

MA ET AL. | 1979

3 | RESULTS

3.1 | Radiolucent mouth retractor

Figure 5 demonstrates the radiolucent properties of the modified radio-

lucent mouth retractor and standard Feyh-Kastenbauer retractor. There

is an appreciable qualitative difference in the visible scatter artifacts.

3.2 | Anatomic tissue displacement

Chart 1 demonstrates the 3D displacement of each oropharyngeal

target as determined between CT scans of the cadaveric heads in the

preoperative versus intraoperative positions. The greatest amount of

displacement was observed in the inferior oropharyngeal targets — the

inferior tonsillar fossa, tongue base, and epiglottis. The highest degree

of displacement was observed in the tongue base (mean 13.77 mm;

range 11.00-16.66 mm), followed by the epiglottis (mean 13.69 mm;

range 10.02-15.75 mm), and the inferior tonsillar fossa (mean 5.53 mm;

range 2.26-9.81 mm). Carotid displacement is also shown in Chart 1

(mean 2.47 mm; range 1.06-5.34 mm).

Figure 6 demonstrates a 3D reconstruction of the oropharynx in the

intraoperative position. The red spheres represent the position of targets

in the preoperative position registered to the intraoperative scan, whereas

the green spheres represent the position of targets in the intraoperative

position. The greatest amount of translation is seen in the inferior oropha-

ryngeal targets— the epiglottis, tongue base, and inferior tonsillar pillar.

3.3 | Tracking system accuracy

Chart 2 demonstrates the TRE of each oropharyngeal target with the

tracking system registered to the preoperative versus the intraoperative

position. A statistically significant reduction in TRE was observed when

registering the tracked pointer to the intraoperative (mouth open) scan

compared with the registration to preoperative imaging. The average

TRE, when registered to the preoperative scan, was 7.63 mm (95% CI

6.79-8.47 mm), whereas registration to the intraoperative scan was

2.49 mm (95% CI 1.95-3.03 mm; P < .05). The greatest TRE was seen

similarly in the inferior oropharyngeal targets (R inferior tonsillar fossa

3.03 vs 6.8 mm; L inferior tonsillar fossa 3.42 vs 6.03 mm; R tongue

base 2.78 vs 14.91 mm; and L tongue base 2.35 vs 13.37; P< .05).

Figure 7 demonstrates a 3D reconstruction of the oropharynx show-

ing tracker accuracy when registered to the intraoperative CT scan.

4 | DISCUSSION

In the present study, we observed that operative positioning for TORS

cases is associated with considerable retraction of the tongue, leading

to significant soft tissue deformation that renders registration to the

preoperative imaging unreliable. In particular, the change in soft

tissue is most pronounced in the inferior aspect of the oropharynx.

Tongue base resections can be especially challenging for a variety

of reasons. The normal lymphoid appearance of the tongue base

mucosa can obscure visualization of tumor extension, whereas a

lack of tactile feedback makes palpating submucosal infiltrating lesions

impossible, and the proximity of the lingual arteries present a real

risk of massive hemorrhage. This is further complicated by the defor-

mation that results from intraoperative positioning, as illustrated in this

study.

TORS has recently been reportedly used for resection of paraphar-

yngeal space masses.10 Many neurovascular structures traverse this

CHART 1 Three-dimensional displacement of each oropharyngeal target between preoperative and intraoperative positioning, in mm. Errorbars represent SD [Color figure can be viewed at wileyonlinelibrary.com]

1980 | MA ET AL.

potential space and can pose considerable risk of hemorrhage. Major

vessel hemorrhage encountered in the robotic setting could be cata-

strophic and difficult to control, given limitations in exposure and

inability to deploy multiple instruments/assistants rapidly. In a case

series by Boyce et al,10 17 patients had parapharyngeal tumors extir-

pated via TORS. The authors report 1 patient with a hematoma requir-

ing reexploration in the operating room for hemostasis. No airway

complications or major hemorrhage were reported. In a cadaveric study

by Wang et al,11 8% of cadaveric specimens demonstrated an abnormal

bulging of the external carotid artery medially toward the superior

pharyngeal constrictor. Although thorough knowledge of the paraphar-

yngeal space is mandatory to perform TORS, advance knowledge of

vascular anatomic variants via intraoperative image guidance can help

surgeons avoid any catastrophic encounters.

One challenge with image-guidance systems is the deformation

that occurs with soft tissue. Although one may achieve good accuracy

when registering to the intraoperative cone-beam CT, these images

remain static with live image guidance in contrast to soft tissue that

gets retracted and manipulated during surgery. In this way, no image-

guidance system is presently able to replicate soft tissue distortion in

CHART 2 Target registration error of each oropharyngeal target with the tracking system registered to the preoperative (mouth closed)versus intraoperative (mouth open) position, in mm. Error bars represent 95% confidence interval [Color figure can be viewed atwileyonlinelibrary.com]

FIGURE 6 Three-dimensional reconstruction of the oropharynx in the intraoperative position. The red spheres represent position oftargets in the preoperative position registered to the intraoperative scan, whereas the green spheres represent position of targets in theintraoperative position [Color figure can be viewed at wileyonlinelibrary.com]

MA ET AL. | 1981

real time via traditional imaging platforms, such as CT or MRI scans. In

contrast, image-guidance systems applied in endoscopic sinus or skull

base surgery are predictable as bony anatomy does not appreciably

deform intraoperatively. Liu et al12,13 and Reaungamornrat et al14 have

published on their image-guidance system using a deformable registra-

tion to link the preoperative and intraoperative images. An intraopera-

tive endoscopic overlay using video augmentation that had been

deformably registered to the preoperative imaging allowed the robotic

surgeon to implant beads into a defined target with increasing accuracy

and reduction in TRE of 5.4 mm.13 Review of the literature reveals a

current paucity of studies in image-guidance in TORS.

There were some limitations to this study. First, we demonstrated

significant reduction in TRE when registered to an intraoperative scan.

However, this may be difficult in many institutions to obtain given the

relative scarcity of intraoperative imaging systems. Conceivably, CT

scans may be obtained with the mouth in an open position, but this

would still not optimally replicate operative positioning and may still be

associated with registration error. Second, the findings of this study are

based on a small series of cadaveric dissections. Last, our custom-

designed radiolucent mouth retractor successfully positions the cadav-

eric specimen to generate an intraoperative cone-beam CT scan with-

out any scatter artifact. However, the tensile strength of the

radiolucent material is far less than the original metal version of the

Feyh-Kastenbauer retractor, requiring larger structural framework of

the mouth retractor components that may be cumbersome for use and

interfere with surgical manipulation. This resulted in a reduction in the

cross-sectional area of transoral access. Nonetheless, to the best of our

knowledge, the radiolucent mouth retractor represents the first retrac-

tor of its kind designed for intraoperative use with image-guidance.

This study demonstrates the quantitative deformation of soft tis-

sue during positioning for TORS and that registration to intraoperative

cone-beam CT imaging significantly reduces TRE. Future work could

look at the application of such a system in a cadaveric tumor model to

evaluate the accuracy and logistical feasibility of image-guidance in

TORS. Similarly, use of injected cadavers could be used to evaluate the

accuracy of navigation around critical vascular structures. At the pres-

ent state, further research is required before an image-guidance system

is ready for clinical use.

As we move into an era of molecular-guided and image-guided sur-

gery, it will be increasingly important to create surgical systems that con-

tinue to use robotic tools in parallel with newly developed imaging and

visualization systems. This article has proven the principle of using radio-

lucent systems with intraoperative near-real time on-the-table imaging,

which has the ability to further enable robotic surgery applications.

CONCLUSION

Considerable tissue translation occurs when a patient undergoing

TORS is imaged in the preoperative position compared to the intrao-

perative position. This image guidance pilot study demonstrates accu-

racy of registration using intraoperative cone-beam CT imaging.

Further work is needed to apply this system in the clinical setting.

ACKNOWLEDGMENTS

The authors would like to thank the Princess Margaret Machine

Shop and Surgical Skills Center at Mount Sinai Hospital, Toronto,

Ontario, Canada. This project was made possible through funding

provided by the Harry Barberian Scholarship, Department of Otolar-

yngology, University of Toronto.

REFERENCES

[1] Hamilton D, Paleri V. Role of transoral robotic surgery in current

head & neck practice. Surgeon. 2016.

[2] Kelly K, Johnson-Obaseki S, Lumingu J, Corsten M. Oncologic, func-

tional and surgical outcomes of primary transoral robotic surgery for

early squamous cell cancer of the oropharynx: a systematic review.

Oral Oncol. 2014;50(8):696–703.

FIGURE 7 Three-dimensional reconstruction of oropharynx showing tracker accuracy when registered to intraoperative CT scan. Thegreen spheres represent position of targets according to intraoperative CT, whereas the red spheres represents the position of the tip ofthe tracked instrument when registered to the intraoperative CT [Color figure can be viewed at wileyonlinelibrary.com]

1982 | MA ET AL.

[3] de Almeida JR, Genden EM. Robotic surgery for oropharynx cancer:

promise, challenges, and future directions. Curr Oncol Rep. 2012;14

(2):148–157.

[4] de Almeida JR, Byrd JK, Wu R, et al. A systematic review of trans-

oral robotic surgery and radiotherapy for early oropharynx cancer: a

systematic review. Laryngoscope. 2014;124(9):2096–2102.

[5] Weinstein GS, O’Malley BW Jr, Magnuson JS, et al. Transoral

robotic surgery: a multicenter study to assess feasibility, safety, and

surgical margins. Laryngoscope. 2012;122(8):1701–1707.

[6] King E, Daly MJ, Chan H, et al. Intraoperative cone-beam CT for

head and neck surgery: feasibility of clinical implementation using a

prototype mobile C-arm. Head Neck. 2013;35(7):959–967.

[7] Daly MJ, Siewerdsen JH, Moseley DJ, Jaffray D, Irish JC. Intraoper-

ative cone-beam CT for guidance of head and neck surgery: assess-

ment of dose and image quality using a C-arm prototype. Med Phys.

2006;33(10):3767–3780.

[8] Daly MJ, Chan H, Nithiananthan S, et al. Clinical implementation of

intraoperative cone-beam CT in head and neck surgery. Proc SPIE

Int Soc Opt Eng. 2011;7964:1–8.

[9] Enquobahrie A, Gobbi D, Turek MW, et al. Designing tracking soft-

ware for image-guided surgery applications: IGSTK experience. Int J

Comput Assist Radiol Surg. 2008;3(5):395–403.

[10] Boyce BJ, Curry JM, Luginbuhl A, Cognetti DM. Transoral robotic

approach to parapharyngeal space tumors: case series and technical

limitations. Laryngoscope. 2016;126(8):1776–1782.

[11] Wang C, Kundaria S, Fernandez-Miranda J, Duvvuri U. A descrip-

tion of arterial variants in the transoral approach to the paraphar-

yngeal space. Clin Anat. 2014;27(7):1016–1022.

[12] Liu WP, Reaugamornrat S, Deguet A, et al. Toward intraoperative image-

guided transoral robotic surgery. J Robot Surg. 2013;7(3):217–225.

[13] Liu WP, Reaugamornrat S, Sorger JM, Siewerdsen JH, Taylor RH,

Richmon JD. Intraoperative image-guided transoral robotic surgery:

pre-clinical studies. Int J Med Robot. 2015;11(2):256–267.

[14] Reaungamornrat S, Liu WP, Wang AS, et al. Deformable image

registration for cone-beam CT guided transoral robotic base-

of-tongue surgery. Phys Med Biol. 2013;58(14):4951–4979.

How to cite this article: Ma AK, Daly M, Qiu J, et al. Intraopera-

tive image guidance in transoral robotic surgery: A pilot study.

Head & Neck. 2017;39:1976–1983. https://doi.org/10.1002/

hed.24805

MA ET AL. | 1983