AbstractFunctional endoscopic sinus surgery (FESS) is a minimal invasive approach adopted in case of chronic sinusitis (inflammation of the paranasal sinuses). The paranasal sinuses are hollow structures within the bones surrounding the nasal cavity. During FESS the surgeon moves the endoscope and other surgical instruments within the nasal cavity following specific paths to approach each one of the paranasal sinuses. The purpose of this study was to reconstruct these paths to access the paranasal sinuses using volumetric CT data. The results will be used for Finite Element modeling and simulations for Robot Assisted Endonasal Surgery.

I. INTRODUCTION VER the past decades, endonasal sinus surgery has become a pathophysiologically oriented, diagnostic-

therapeutic general concept. Functional endoscopic sinus surgery (FESS) is a minimal invasive approach adopted in case of inflammation of the sinuses [1]. The surgery restores physiology by reestablishing mucociliary drainage and ventilation of the sinuses. When treating a chronic sinusitis, for example, a standard management system consisting of complementary partial steps, specific combination depends on the extent of mucosal disease present [2].

Several improvements related to FESS have been adopted on an international scale and expanded. More comprehensive statistics have led to important insights as to the type and incidence of complications caused by endonasal surgery and as to the special therapy indicated. These endonasal procedures and minimal invasive techniques attempt to correct a diseased tissue site, to minimize the risks involved in surgery, to achieve a reduction of pain, and to reduce the duration of hospitalization periods [3]. Furthermore, there is a large spectrum of indications that use the nasal cavity and the paranasal sinuses to access others structures. The cases are falls of chronic mucositis, trauma, foreign body and extraction of tumors [4-6]. A. Nasal Cavity and Paranasal Sinuses The nasal cavity is a large air-filled space above and behind the nose in the middle of the face. It is divided into right and left halves by the nasal septum. Four pairs of hollow

Manuscript received June 22, 2007. A.I. Moral, M.E. Kunkel, M. Rilk and F.M. Wahl are with the Institute of Robotics and Process Control, Technical University Braunschweig, Germany (phone: 0049531-391-7450; e-mail: eku@rob.cs.tu-bs.de). K. Tingelhoff, I. Wagner, K. W. G.Eichhorn and F. Bootz are with the Clinic and Policlinic of Otolaryngology/Ear, Nose and Throat Surgery, University of Bonn, Germany (phone: 0049228-2871-5552; e-mail: ktingelh@ukb.uni-bonn.de).

structures within the bones surrounding the nasal cavity form the paranasal sinuses. The sinuses are divided into subgroups named according to the bones they lie under. The frontal sinus is placed over the eyes, in the forehead bone; the maxillary sinus is placed, under the eyes, in the upper jawbone; the ethmoidal sinuses are comprised of a variable number of air cells, ranging from 3 to 18 on each side, and they are between the nasal septum and the eyes, backwards into the skull; and finally, the sphenoidal sinuses are placed in the center of the skull base [7,8]. Computed tomography (CT) is currently a standard procedure used for two-dimensional representation of the nasal cavity and paranasal sinuses [9]. B. Purpose of this work The goal of the current study was to reconstruct the paths to access the paranasal sinuses using volumetric CT data. This investigation is part of the development of automated robotic endoscope guidance for FESS [10,11]. Such a technological innovation requires accurate understanding of the dynamics of the endoscope related to the spatial distribution of the nasal cavity structures.

II. MATERIAL AND METHODS To obtain the three-dimensional reconstruction of the head and paranasal sinuses, two CT datasets acquired by a spiral CT from Philips were used. The first dataset consists of 486 CT images with resolution of 0.0664063 x 0.0664063 x 0.05 mm from a cadaver head. The second dataset consists of 98 transversal slices, each 1.0 mm thick, with a resolution of 512 x 512 pixels. The pixel spacing is 0.346 mm x 0.346 mm. This second dataset was acquired from a female patient age 37. It shows little chronic maxillary sinusitis (bilateral) and bullose middle turbinate (bilateral). Furthermore, the patient has unaffected paranasal sinuses.

Semi-automatic segmentation of the images was performed using AMIRA 4.1 (Mercury Computer System, Inc., USA) [12]. The visualization of the CT images was based on a user defined Hounsfield windows of 2000. The semi-automatic segmentation was performed by a 3D growing region algorithm, which starts from a manual chosen seed point (Fig.1). Then the largest connected region is segmented with voxels whose gray values lie inside a user defined range. Values greater than -200 were used to segment soft and hard tissues, while for paranasal sinuses and nasal cavity values smaller than -600 were used.

Due to the complicated geometry of the nasal cavity and paranasal sinuses, the segmentation was performed slice per slice and that consumed long time periods [13]. On the other

3D Endoscopic Approach for Endonasal Sinus Surgery A.I. Moral, M. E. Kunkel, K. Tingelhoff, M. Rilk, I. Wagner, K. W.G. Eichhorn,

F. Bootz and F. M. Wahl

O

Proceedings of the 29th Annual InternationalConference of the IEEE EMBSCit Internationale, Lyon, FranceAugust 23-26, 2007.

SaP1C5.6

1-4244-0788-5/07/$20.00 2007 IEEE 4683

hand, the segmentation of the entire head, including soft and hard tissues can be done in one single step. After segmentation a triangular approximation of the interfaces between air and nasal cavity mucosa was computed.

Fig 1. Computed tomography in coronal slice orientation of a patient head showing the segmentation of the nasal cavity and paranasal sinuses. Maxillary sinuses (M), ethmoidal sinuses (E) and nasal cavity (NC). The arrow indicates the seed point and the region calculated by the 3D growing algorithm.

III. RESULTS After the segmentation, the approach to all paranasal sinuses (maxillary, ethmoidal, sphenoidal and frontal) was reconstructed, using to that purpose landmarks. These landmarks were settled starting on the nostrils and ending in the natural openings of the sinuses. The placement of the markers can be done, either in the CT images or in the three-dimensional views.

A. Endoscopic Endonasal Approach to the Maxillary Sinuses The access to the maxillary sinuses is physicologically by the middle nasal passage (supraturbinal). The opening in the inferior nasal passage (infraturbinal) is not physiological and should not be used any more (Fig. 2).

Fig. 2. The paranasal sinuses are represented in transparent visualization of a three-dimensional reconstruction of a cadaver head. The structures are the maxillary, ethmoidal, frontal and sphenoidal sinuses. The spheres (landmarks) show the access from the nostril to the maxillary sinus.

Under endoscopic vision the middle turbinate is medially moved and the thin wall of the maxillary sinus underneath the processus uncinatus is opened [13]. The passage to the orbital floor communicates superiorly with the limit of the opening for the maxillary sinus.

B. Endoscopic Endonasal Approach to the Ethmoidal Sinuses The ethmoidal cells can be accessed starting from middle nasal passage. The problems of this access are regarded the variations found in the anatomy of the ethmoidal roofs [8]. When the surgeon has to penetrate deeply a lesion in the skull basis can occur (Fig.3).

Fig. 3. Lateral three-dimensional visualization of the access route from the nostril to the ethmoidal sinus in a cadaver head. The structures are the maxillary, ethmoidal, frontal and sphenoidal sinuses. The spheres (landmarks) show the access to the ethmoidal sinuses.

C. Endoscopic Endonasal Approach to the Sphenoidal Sinuses The pathway to the sphenoidal sinuses runs oblique through the nasal cavity, between the nasal septum and the middle turbinate. The spheres in Figure 4 indicate the free way from the sphenoidal sinus, through the nasal cavity. In this case, the points were marked in only one slice. Figure 5 shows a more realistic pathway to the sphenoidal sinus, which suggests that the endoscope has to push and shift the turbinates in order to reach its aim. Once the endoscope reaches the interior of the sphenoidal sinus the most dangerous complications are due to the damage of the wall of the carotid and the optic canal.

nostril

nostril

M M

E E

NC NC

seed point

Frontal Sinus

Maxillary sinus

Maxillary sinus

Ethmoidal sinuses

Frontal Sinus

Maxillary sinus

Ethmoidal sinuses

Sphenoidal sinuses

4684

Fig. 4. Axial view of CT image from a patient head. The spheres indicate the free way to the sphenoidal sinus, through the nasal cavity.

Fig. 5. Lateral three-dimensional visualization of the access route from the nostril to the sphenoidal sinuses in a patient head. The structures are the maxillary, ethmoidal, frontal and sphenoidal sinuses. The spheres (landmarks) show the access to the sphenoidal sinus.

D. Endoscopic Endonasal Approach to the Frontal Sinuses The right and left nasal apertures of the frontal sinus are always separated by the nasal septum. A variable number of ethmoidal cells may reach the frontal sinus, but only two of them will initiate the sinuses. The other will complete for space, remaining at the base of the sinus or extending elsewhere. The common sites of opening of the frontal sinus in the nasal cavity are the infundibulum, frontoethmoidal recess, in front of the semilunar hiatus, in its anterior end, in its anterior one fourth, above the semilunar hiatus, in the frontal recess, above the superior boundary of the infundibulum, in the infundibulum, in the infundibular region[8]. Figures 6 and 7 show the access route from the nostril to the frontal sinuses in a patient head. To obtain this path a diagonal line was marked between the nostril and the inferior opening of the frontal sinuses with opens to the ethmoidal sinuses.

Fig. 6. Frontal view of partial three-dimensional reconstruction (left) and coronal view of CT image (right) of a patient head. The paranasal sinuses are represented in transparent visualization of a three-dimensional reconstruction of a patient head. The structures are the maxillary sinuses (MR and ML), frontal sinuses (FR and FL) and nasal cavity (NCR and NCL). The spheres (landmarks) show the access from the nostril to the frontal sinus (FL).

Fig.7. Lateral three-dimensional visualization of the access route from the nostril to the sphenoidal sinuses in a patient head. The structures are the maxillary, ethmoidal, frontal and sphenoidal sinuses. The spheres (landmarks) show the access to the sphenoidal sinus.

IV. DISCUSSION AND CONCLUSION The 3D reconstruction of the endoscopic endonasal access routes to the paranasal sinuses were performed from CT images of cadaver head and a patient. Four approaches to the maxillary, ethmoidal, sphenoidal and frontal sinuses were studied. During the segmentation the connections between the paranasal sinuses and the nasal cavity were identified, which permitted us to determine the path to access each sinus individually during FESS. In previous works we have investigated the force required to access to the maxillary, frontal and ethmoidal sinus. No data about the shpenoidal sinus are available. The

Septum

Ethmoidal sinuses

Sphenoidal sinuses

4685

measurements took place in five patients, in two of them only ethmoidal and maxillary sinuses were accessed. The results showed that in average the biggest forces were required to access the maxillary sinuses (3.72 N), while the smallest were required to access to the ethmoidal sinus (2.58 N). For the frontal sinus an average of 2.93 N was required. The great inconvenient of these data is that only five cases where investigated, but on the other hand, these results are in good concordance with the path to follow for the endoscope to reach the sinuses. That is, the ethmoidal sinuses, which encountered the minimal average force, are reached just in the roof of the nasal cavity, and no strength ostia have to be overcome to reach its inner. However, frontal and maxillary sinus, whose entrances are more obstructed showed maximal forces. In this way, greater forces are hoped for the shpenoidal sinuses, in whose case the endoscope has to travel along the nasal cavity, and furthermore, to open and traverse ostia.

Although not included in this article, the coordinates of the landmarks in the CT images can be known. These measures are important to describe a safe trajectory to follow by an endoscope [10,11]. For us these results are of great importance, since we try to determine which structures are more affected due to the contact with the endoscope. Some studies show that this kind of visualizations not only are helpful to prevent accidents, to plan the surgical procedures, diminish the operation time, etc but also help the surgeon to improve its technique [4].

In future works we will also determine the points of contact between endoscope and the structures. For this purpose we are looking forward to correlate the images of this work with videos obtained during live operations. We have also recorded the position of the tools using a tracking system and the forces exerted by the surgeon when he was trying to access to the paranasal sinuses. All these data will give us a detailed description of the deformation that the nasal cavity structures suffer during FESS, und thus to determine which ones are subject of risk and to prevent accidents.

In the next step we will use this information to simulate the biomechanical behavior of the nasal structures under endoscope loading during FESS. Other important applications of these results are the geometrical model of the entire head for trajectory planning, the study of anatomic variations of human heads and simulations for medical training.

ACKNOWLEDGMENT

The authors are grateful to Prof. Dr. K. Schild and Priv.-Doz. Dr. med. Wilhelm, Clinic of Radiology, University Bonn for providing CT image data and to Prof. Dr. K. Schilling and Dr. I. Kinsky, Department of Anatomy, University Bonn for providing the cadaver head. This work is part of the project Robot assisted intuitive endoscope navigation in endonasal surgeries with the help of preoperative CT or MRT analysis and we are thankful that

the Deutsche Forschungsgemeinschaft (DFG) funds this project.

REFERENCES [1] L.F. Smith, P.C, Indications, evaluation, complications, and results

of functional endoscopic surgery in 200 patients, in Otoryngol Head Neck Surg, 108, 1993, pp. 688-696.

[2] S.J. Zinreich, D.W. Kennedy, A.E. Rosenbaum et al. Paranasal Sinuses: CT Imaging Requirements or Endoscopic Surgery, in Radiolog. 163, 1987, pp.769-75.

[3] H. Iro and W. Hosemann, Minimally invasive surgery in otorhinolaryngology, in Eur. Arch. Otorhinolaryngol, 250, 1993, pp. 1-10.

[4] G Widmann, "Image-guided surgery and medical robotics in the cranial area," in Biomedical Imaging and Intervention Journal, 2007.

[5] Sau-Tung Chu, Endoscopic Sinus Surgery Under Navigation. System-Analysis Report of 79 Cases, in J Chin Med Assoc, vol 69, No 112006, 2006..

[6] H.L. Levine, Functional endoscopic sinus surgery: evaluation, surgery, and follow-up of 250 patients, in Laryngoscope. 100, 1990, pp. 79-84.

[7] J. Lang, Clinical Anatomy of the nose, nasal cavity and paranasal sinuses, Thieme Verlag, NY:New York, 1989.

[8] J.A.C. Navarro, The nasal cavity and paranasal sinuses: Surgical Anatomy. Springer-Verlag, Berlin:Germany, 2001.

[9] H Shi, W.C Scarfe, A.G. Farman, Maxilary Sinus 3D Segmentation and Reconstruction from Cone Beam CT Data Sets, in Int J CARS. 1, 2006, pp.83-89.

[10] K. Tingelhoff, I. Wagner, K.W. Eichhorn, M Rilk, R. Westphal, F.M. Wahl, F. Bootz, Sensor-based force measurement during FESS for robot assisted surgery. GMS CURAC, vol. 2, no. 1, 2007.

[11] M. Rilk, S. Winkelbach, F. Wahl, Partikelfilter-basiertes Tracking chirurgischer Instrumente in Endoskopbildern, in Bildverarbeitung fr die Medizin, 2006, pp. 414-418

[12] Mercury Computer System, Inc., USA. Amira Reference Guide. [13] A.I. Moral, M.E. Kunkel, M. Rilk, F.M. Wahl, K. Tingelhoff, F.

Bootz. Paranasal sinuses segmentation/reconstruction for robot assisted endonasal surgery, Aceppted at Thematic Conference on Computational Vision and Medical Image Processing ECCOMAS VIPIMAGE, 2007.

[14] W.E. Thumfart, W. Platzer, A.R. Gunkel, H. Maurer, E. Brennen, Operative Zugangswege in der HNO-Heilkunde. Thieme Verlag, Stuttgart:Germany, 1998

4686

MAIN MENUGo to Previous DocumentCD-ROM HelpSearch CD-ROMSearch ResultsPrint