A, A bracket with a simple shield of infrared diodes. B, The operating microscope equipped with the bracket holding a shield of 4 infrared diodes that must be detected by the optical tracking system.
Operating microscope focal point calibration tool, which is also equipped with a simple shield of 4 infrared diodes that are detected by the optical tracking system. Of 2 mechanically aligned crosshairs, the center of the bottom crosshair is selected as the calibration point. Each time, the operating microscope focuses on this point through the center of the top crosshair so that in the microscope view, the center of the top crosshair is superimposed onto that of the bottom crosshair.
The 11 landmarks on the anterior skull base that are used for measuring system accuracy.
The 11 landmarks on the lateral skull base that are used for measuring system accuracy.
Results of testing the microscope's accuracy in the anterior skull base. Each landmark was tested 20 times.
Results of testing the microscope's accuracy in the lateral skull base. Each landmark was tested 20 times.
Zheng G, Caversaccio M, Bächler R, Langlotz F, Nolte L, Häusler R. Frameless Optical Computer-Aided Tracking of a Microscope for Otorhinology and Skull Base Surgery. Arch Otolaryngol Head Neck Surg. 2001;127(10):1233-1238. doi:10.1001/archotol.127.10.1233
To integrate a digitally controlled operating microscope without a laser autofocus system into a frameless optical computer-aided surgery system and to test the accuracy and usability of this system in otorhinological surgery.
Experimental study and case series.
Department of Oto-Rhino-Laryngology, Head and Neck Surgery, Inselspital, and the Maurice E. Müller Institute for Biomechanics, University of Bern, Bern, Switzerland.
Eight computer-aided microscopic surgical procedures were performed between January and October 2000 on patients with various diseases of the anterior and lateral skull base.
The practical accuracy of the navigated microscope on the lateral side of a cadaver skull was 2.27 ± 0.25 mm and on the anterior side of the same skull was 2.07 ± 0.35 mm. In all 8 cases of computer-aided microscopic surgery, no complications occurred. Clinical inaccuracy was 2 to 3 mm.
Integration of a low-cost, non–laser autofocus microscope into our computer-aided surgery system was successfully performed and offers surgeons the ability to combine the precise optics of the operating microscope with the localization power of a computer-aided system.
THE TECHNOLOGY for computer-aided surgery (CAS) and indications for its use in otorhinolaryngologic (ORL) practice have emerged in the past 10 years.1- 3 Ultrasonic, electromechanical, electromagnetic, and optoelectronic guiding systems have been proposed to support the surgeon's 3-dimensional intraoperative orientation and to prevent complications. These systems allow the surgeon to reach targets without direct visual control. Combined with anatomical information obtained from magnetic resonance imaging, computed tomography (CT), and angiography, CAS offers a safe and minimally invasive approach to the treatment of skull base lesions. In general, 3 major components are common to CAS systems. (1) The real object is that part of the anatomy to be approached surgically. In the case of ORL surgery, this is the skull. (2) A virtual object is any kind of image representing the real object and allowing for better insight into its interior. Computed tomographic scans are commonly used for this purpose in ORL surgery, but magnetic resonance images and angiography may serve as the virtual object as well. (3) The navigator is an aiming device that guides the surgeon to a point in the real object that was defined in the virtual object.
Today, most of the demanding procedures in ORL surgery require the aid of the operating microscope. In microsurgery, in which use of a specially designed operating microscope allows for the performance of controlled tissue removal under improved lighting and magnification, there is a great need for online information about patient anatomy and the orientation and position of a surgical tool. However, the microscope and handheld navigation instruments cannot be used at the same time.4 Because of the size of the microscope and its sterile cover, the space between the instruments and the optoelectronic digitizer is shadowed. To measure without removing the microscope from the surgical field, it must be used as a localizing instrument itself.4
In this article, we propose an effective method for integrating a digitally controlled microscope into a frameless optical CAS system in a theoretical and practical trial. A bracket with light-emitting diodes (LEDs) is attached to the microscope, and a specially designed calibration tool is used to calibrate the microscope's focal point precisely. During surgery, the focal plane and trajectory of the microscope are tracked by the optical tracking system in real time, and the information is correlated with the preoperative scan images and updated interactively. The intraoperative video image from the operating microscope is displayed side by side with cropped slices of preoperative scan images.
Informed consent was obtained from all patients; however, the study was classified as exempt by the local institutional review board.
The CAS system (SurgiGATE ORL; Medivision, Synthes-Stratec Medical, Oberdorf, Switzerland), developed by a research group in Bern, Switzerland, is based on CT scans.5 The images are saved in a digital archiving system by means of a network or are directly transferred to the computer in the operating room (Sun Workstation [Unix]; Sun Microsystems, Mountain View, Calif). The frameless optical tracking system used in the operating room included an optoelectronic space digitizer (OPTOTRAK 3020; Northern Digital Inc, Waterloo, Ontario) as the passive navigation device. The precalibrated system can track up to 256 pulsed infrared LEDs with an accuracy of 0.1 mm. Three or more LEDs can be defined to represent a rigid body to be followed globally with respect to the camera's frame of reference or locally with respect to any other rigid body. Each surgical instrument carries a shield with 4 LEDs to form a rigid body that defines the tracked instrument's pose (position and orientation) and its graphical representation through points of interest (eg, the tool tip and the tool axis). This optical tracking system allows the surgeon to get online spatial and geometrical information about any surgical instrument in use.
Matching, or registration, is defined as the transformation between the real and virtual objects and is a critical aspect in most CAS applications. Two different methods were developed and evaluated by our group6: (1) a paired point–matching algorithm, which determines the relation between both objects by a number of discrete points given in both worlds, and (2) a surface-matching algorithm that does not use any fiducial markers. This registration method uses 15 to 18 points captured by the surgeon with a space pointer to represent the patient's anatomical features. The first 3 points correspond to anatomical landmarks defined in the CT images, eg, the frontozygomatic suture, the nasion, and the anterior nasal spine for the anterior skull base. These so-called coarse landmarks do not need to be captured precisely—as much as 20 mm of error is accepted. The remaining points lie on the bony surface of the skull and are referred to as surface points. The coarse landmark pairs are used to get an initial estimate of the transformation and to restrict the matching algorithm to the area of the correct solution. The surface points are used to search a fine solution of the transformation.
In our study on the lateral skull base, we recommended that the following 5 anatomical points be selected for paired point matching with our navigation system: the tip of the mastoid, the mastoid foramen, the umbo, the frontozygomatic suture, and the anterior nasal spine. For additional accuracy in clinical situations, surface matching is recommended.6
During an image-guided intervention, it is necessary to compensate for any relative motion between the patient and the localizer. If this were not done, then the registration would become invalid as soon as either the patient or the localizer moved. For marker-based surgical instrument localization, the common technique to achieve referencing is the attachment of a dynamic reference base to the patient's anatomical features before registration. The dynamic reference base establishes a local frame of reference for the bone to be operated on. With the aid of this device, the localizer can collect the position data of navigated instruments with respect to the anatomical features, thus neutralizing eventual patient and localizer motion. In our computer-aided microscopic surgery system, the dynamic reference base is positioned on the head of the intubated patient. It consists of an upper jaw (maxilla) splint equipped with a rod and is attached to the jaw by means of silicone impression material of medium consistency (Coltène, Altstätten, Switzerland). The rod is equipped with LEDs. If the patient is edentulous, then the same splint with a silicone mass is needed. In addition, we fix the splint using 1 to 2 miniscrews (diameter, 2.0 mm; length, 12-14 mm) (Synthes-Stratec Medical) on the hard palate.5 The miniscrews are left on until the end of surgery.
The operating microscope used in this work consists of a digitally controlled operating microscope and a floor stand (models VM900 and FS3013, respectively; Möller-Wedel GmbH, Wedel, Germany). It provides a communication interface for navigational systems to read or send commands and variables about all movements of the operating microscope, such as the moving distance of the focal point and the zooming scale. A camera is used to capture the video image. The video output of the camera is connected to a television monitor and to a standard frame grabber (MultiMedia Access Corporation, Cary, NC) installed in the workstation.
To integrate the digitally controlled operating microscope into a CAS system, the following specifications have been established:
Conforms to the cleanliness and sterilization standards for equipment in the operating room.
Does not interfere with normal operative procedures or limit the surgeon's access to the operative field.
Does not interfere with the surgeon's use of the CAS system with all other existing CAS tools.
As a new CAS tool, should require minimum alteration of the existing CAS system and should be compatible with other CAS tools.
Requires minimum alterations in the operating microscope and its floor stand, none of which should void the microscope's warranty.
To achieve these objectives, we developed a special bracket that is mounted on the operating microscope frame and that holds a shield with 4 LEDs (Figure 1). In designing this device, care was taken so that the operating microscope was attached as rigidly as possible. At the same time, the shield must not be occluded by any part of the operating microscope and the floor stand so that the 4 LEDs can always be seen by the optical tracking system. These LEDs establish a local coordinate system for the microscope that defines the operating microscope as a traceable rigid body. This computer-aided microscopic system can be guided by a virtual keyboard.7
At a particular zooming scale, the focal point in any position is rigidly fixed to the frame of the operating microscope. Therefore, after rigidly attaching a shield with 4 LEDs on the frame of the microscope to establish a local coordinate system for the microscope, it is possible, through a calibration procedure, to find the position of the focal point in this coordinate system. Later, with the help of the optical tracking system, we can transform this position to the coordinate system defined by the dynamic reference base and finally to the image space with a matching transformation.
A special tool has been designed to calibrate the operating microscope (Figure 2). The basic principle is to solve a linear equation between the steps that the focal point has moved and the coordinate of this focal point in the local coordinate system of the microscope. Later, no matter where the focal point is and based on the steps that the focal point has moved, we can calculate its corresponding spatial coordinate in this local coordinate system based on the linear equation.
During calibration, in nonspecified varied distances between the microscope and the calibration tool, we focus the microscope on the center of the bottom crosshair of the calibration tool through the center of the top crosshair. The position of the bottom center in the local coordinate system of calibration is known before calibration. We retrieve the steps each time that the focal point has moved from the microscope itself and the coordinates of the bottom center in the microscope's local coordinate system from the optical tracking system. Based on these data, we can calculate the coefficients that define the linear relationship.
Accuracy measurements were performed on a cadaver skull model and in 8 patients undergoing ORL microsurgery for a variety of diseases. The euclidean distance was used. Microsoft Excel 97 (Microsoft Corp, Redmond, Wash) was used for evaluation of the data.
A needle pointer with a tip diameter of 0.5 mm was used. The calibration unit defines the point of interest (eg, the tool tip, axis, and orientation) accurately and checks the tool geometry by a simple point comparison. We used this pointer and the calibrated microscope to digitize the same landmark point on the skull model. System accuracy was assessed by measuring the deviations between the digitized results, where one is the output from the needle pointer and the other is based on the microscope. Two sets of points were chosen as measuring points from different anatomical areas of the skull (Figure 3 and Figure 4) that correspond to areas that commonly undergo ORL surgery. One is on the anterior skull side and the other is on the lateral skull side.
In our preliminary clinical evaluation, 8 patients were treated on the anterior and lateral skull base (for recurrent polyposis nasi [n = 3], frontal recess stenosis (n = 2), osteoma of the frontal recess [n = 1], or ear malformation [n = 2]) with our computer-aided microscopic surgery system based on paired and surface matching. During surgical planning, either an axial CT scan of the skull (sectional thickness, 1.5 mm; spacing, 1.5 mm) or a helical scan was performed. Reliable anatomical paired points were then marked on the CT scan, such as in the case of the anterior skull base, the frontozygomatic sutures, the nasion, and the anterior nasal spine. Neither fixation of the head with frames nor insertion of invasive markers is needed. In addition, possible surgical trajectories (paths to the target and goals) can be identified.
During microsurgery, after focusing the microscope on an anatomical area, the needle pointer was used to point to the same position. The positions of the microscope's focal point and the needle pointer tip were transformed to image space, and the deviation between the 2 points in image space was calculated. The deviation was then measured in multiples of the pixel size of the CT images.
The deviation errors for points in different anatomical areas are shown in Figure 5 and Figure 6. In the lateral assay, a mean ± SD error of 2.27 ± 0.25 mm (range, 1.96-2.71 mm) was determined. The upper confidence limit (P>.95) was 2.76 mm. In the anterior assay, the mean ± SD error was 2.07 ± 0.35 mm (range, 1.77-3.07 mm), and the upper confidence limit was 2.76 mm.
We also observed that although the mean error on the lateral skull base was slightly larger than that on the anterior skull base, the SD on the lateral skull was slightly smaller than that on the anterior skull. This is probably because the landmarks selected on the lateral skull base were more densely located in almost the same plane, whereas the landmarks on the anterior skull were more sparsely distributed spatially in different planes. This also explains why the maximal value on the anterior skull side was larger than that on the lateral skull side, although the mean value on the lateral skull side was smaller. This point has been verified by our clinical testing.
In all 8 patients who underwent computer-aided microscopic surgery, no complications occurred. However, a problem remains, ie, the focal level in the case of bleeding, in which the deviation error is higher (>3.0 mm).
Several methods have been proposed previously for integrating the operating microscope into the CAS system. Different technical approaches have been used clinically, such as passive mechanical arms,8 active robotic arms,9 and ultrasonic,10 electromagnetic,11 and optical digitizing4,12,13 systems. Currently, 3-dimensional digitizers based on optical sensor technology are supposed to be the most accurate systems because they are less affected by disturbing environmental factors than are the others. Furthermore, in contrast to mechanical arms, they allow an unobstructed view of the surgical site and a straightforward surgical procedure in the operative field.
Our proposed method is somewhat similar to that used by Hauser and colleagues4,12,13 in that we both used an optical digitizing system to track the microscope. However, the main difference between the system used by Hauser and colleagues and ours is that a stereotactic frame is required in their system as a patient registration and reference system, which is often disadvantageous for ORL surgery. In their system, the frame is mounted transitorily 1 day before surgery for CT acquisition. On the day of surgery, the frame is repositioned on the patient's head. Our system is frameless. We use software technology to solve the registration problem. In addition, in their system, the shield holding the LEDs is mounted to the housing of the microscope, and an object distance-measuring unit is attached in front of the objective lens. In our system, the shield holding the LEDs is mounted to the frame of the microscope, and no other unit is required. The moving distance of the focal point is retrieved from the microscope itself.
Compared with other surgical instruments, the non–laser autofocus microscope is less accurate. In our previous study,5 the practical accuracy for other surgical instruments on the cadaver skull was 0.5 to 1.2 mm, whereas the mean ± SD error for the microscope was 2.07 ± 0.35 mm on the anterior skull and 2.27 ± 0.25 mm on the lateral skull. These larger spatial errors can probably be attributed to the following factors: (1) The instrument's trajectory is defined by its origin and tip as the rigid end of the longitudinal axis. In contrast to the geometrically predefined tip of the instrument, the microscope's focal point and associated optical trajectory must be localized by surgeons because an operating microscope without a laser autofocus system is used. Different surgeons will have different focal levels. The large range of individual visual adaptation of the surgeons makes it difficult to determine the exact focal point of the microscope. (2) Errors can be introduced directly by the microscope and are referred to as errors caused by the optical properties of the microscope, such as the method of focal point adjustment, focal length, and magnification of the lens. (3) Errors can be introduced indirectly by the microscope and are referred to as errors caused by tracking properties of the microscope, such as the tracking distance, the number of the reference infrared diodes, and the area covered by these infrared diodes. (4) The condition when intraoperatively tracking the microscope can be different from the condition when calibrating the microscope, such as microscope orientation, tilt, or rotation. (5) The anatomical structures in the microscopic surgical site can affect accuracy, in that the more spatially sparse the distribution of the anatomical structures in the surgical site, the less accurate the microscope.
Clinically, we found that a problem remains. The inaccuracy of the focal level is greater than 3 mm in the case of bleeding. Because of the well-vascularized mucosa of the nose, it is common to find such a situation. On the lateral skull base, the soft tissue is not as thick as in the nose, except in chronic inflammation. In addition, the bleeding is usually less than in the nose, except for in well-vascularized tumors. This means that microscopic navigation on the bone of the lateral skull base with an exact navigation level of the focal point can be more accurate in clinical practice. The handling of the navigation microscope with only a shield of infrared diodes (Figure 1) is not impaired. The Möller-Wedel microscope is a low-cost system (US $40 000) compared with other microscopes. However, it has the disadvantage of being a non–laser autofocus system.
In conclusion, a new method has been proposed to successfully integrate a low-cost non–laser autofocus operating microscope into a CAS system without interfering with the surgical procedure and changing any existing CAS tools. It offers surgeons the ability to combine the precise optics of the operating microscope with the localization power of the CAS.
Accepted for publication June 27, 2001.
Presented in part at the 4th European Congress of Oto-Rhino-Laryngology, Head and Neck Surgery, Berlin, Germany, May 18, 2000.
Corresponding author and reprints: Marco Caversaccio, MD, Department of Oto-Rhino-Laryngology, Head and Neck Surgery, University Hospital, Freiburgstrasse, CH-3010 Bern, Switzerland (e-mail: firstname.lastname@example.org).