The challenge of endo- and transnasal surgery
Endonasal surgery refers to surgery of the nasal cavities and sinuses, while transnasal surgery here refers to surgery using an endonasal approach to anatomic structures lying beyond the nasal cavities and paranasal sinuses: infratemporal fossa, endo-orbital structures anterior and ventral skull base, and petrous apex.
In current practice, endonasal surgery remains strongly affected by its origins: Messerklinger, Agrifolio and Terrier were modest in their aims, however innovative, seeking to approach the maxillary and sphenoid sinuses and ethmoid labyrinth using the recently developed panoramic endoscope. Despite the advantages of this new technique ( Figure 10.1 ), many teams long remained faithful to the operative microscope, especially for sphenoid and ethmoid surgery, as it enabled two-handed surgery. However, as surgeons mastered the technique, transnasal surgery came gradually to be extended to regions hitherto requiring external neurosurgical approaches, sometimes involving craniotomy and its sequelae. In France, Jankowski [ ] was probably the first to see that endonasal surgery had possibilities going beyond mere rhinology; he described a strictly naso-sphenoidal sellar region approach, opening the way to the advent of rhinoneurosurgery. Other teams, including those of Castelnuovo and Nicolai [ ] in Italy, Kassam, Carau and Snyderman in the United States [ ], and Herman in France [ ], realized and developed the possibilities of such surgery, pushing the bounds, refining indications and teaching the technique.
Despite tremendous progress in this field, endonasal surgery mostly relies upon the conventional instrumentarium of its earliest days, with few changes so far, the only noticeable exception being the advent of surgical navigation.
In this chapter, we will try to see how robotized systems could improve endonasal surgery and what attempts are being made in this direction.
What to expect from a robotized manipulator in endonasal surgery
As a general rule, robots are expected to be more accurate and possibly faster than surgeons. In endo- and transnasal surgery, some specific reasons suggest that a robot could improve the way that this surgery is currently performed.
Improving surgical workflow by restoring two-hand surgery
Analyzing the gestures of senior surgeons, even those highly skilled in this surgery, shows that a significant part of the procedure time is spent on tasks not directly consisting in dissecting or resecting tissue to be removed or repaired. Analyzing videos of real procedures ( Figure 10.2 ) shows that 20% to 50% of overall surgery time is spent aspirating accumulated blood to allow visualization of the surgical field, washing the field with water, and trying to find the right-shaped instrument for anatomical structures that are difficult of access. Another frequent difficulty is the impossibility of distracting tissue before cutting, a requirement obvious to any dressmaker. There is obviously much room for improvement in endo-transnasal surgery workflow.
It is strongly suspected that most of these limitations could be overcome by allowing the surgeon to use both hands rather than working single-handed. And it is reasonable to expect that improving surgical workflow could reduce total operating time.
Management of the introduction and removal of surgical instruments
The insertion of instruments inside the nasal tract is a tedious task that has to be done hundreds of times during a procedure. Each introduction can inflict trauma on the nasal mucosa, especially in the nasal vestibule, which usually cannot easily be seen directly or endoscopically. It is guessed that a surgical manipulator could speed up the introduction and removal of instruments while at the same time constantly controlling optimal orientation to minimize unwanted contact with the patient’s nasal tract.
Managing several instruments simultaneously
While most open conventional surgeries use many instruments simultaneously for retracting, dissecting and sucking, conventional endonasal surgery is usually performed using a single instrument at a time, manipulated by the surgeon’s dominant hand, while the non-dominant hand is devoted to endoscope manipulation. However, simultaneous use of several instruments is a powerful means to enhance surgical workflow, improve tissue dissection, and maintain a clean surgical field even when drilling bone.
Some authors suggest using both hands, with an assistant surgeon holding the endoscope [ ]. Although justified for sophisticated 2-surgeon 2-nostril procedures, this technique is unsuited to more common surgeries such as ethmoidectomy. Manipulating the suction cannula and forceps in the same hand has been suggested [ ], but with obvious limitations. Some static arms are available, with various locking technologies to hold endoscopes; but, being immobile once secured, they cannot adapt to the course of co-working instruments. Further, due to their static nature, they cannot move out of the endoscope when instruments are extracted from the nostril, and thus require special endoscopes with extended length to allow the normal manipulation of forceps [ ]. This makes them useful only in certain contexts, such as resection of a pituitary adenoma by a trans-sphenoidal approach.
Another significant expectation of robots is the possibility of letting them find the best possible instruments orientations so as to increase the number of those usable via the same nostril ( Figure 10.3 ). This is merely vector calculus, for which human beings have poor capabilities, unlike numerical systems which can very quickly and in real time compute the best solution if the anatomical space, instrument geometries and their current positions are known.
Navigation-based endonasal robotics
Except for neurosurgical transcranial approaches [ ], navigation has rarely been linked to robotics despite the very promising enhancement such an association could provide. We believe that endo- and transnasal robotic surgery is a perfect candidate for such combined technologies, providing the robot controller with a map so as to define which prerequisites have been to be taken into account to optimize safety and minimize trauma for the patient. This concept is exactly the same as the flight plan used in aerospace contexts.
Once a trajectory solution is provided by the computer, some spatial limits may be added as virtual fixtures obliging the robot to adjust its motion with respect to these boundaries with adjustable levels of priority: free area with maximum velocity allowed, unsafe area with limited speed, strictly forbidden areas.
While automated tissue segmentation ( Figure 10.4 ) is in a mature state [ ], the main current limitation to the development of this technology is our limited knowledge of how to model endonasal soft-tissue elasticity efficiently. Using the finite elements method (FEM), some studies of a simplified 3D biomechanical model to estimate forces and torques induced by deforming nasal tract soft tissues have been published, but no clinical evidence of the reliability of such a method has yet been demonstrated [ , ].
While augmented reality has penetrated many fields such as the aerospace, automotive, military and games industries, ENT endoscopy, despite significant progress in imaging quality, remains at the same conceptual level as 60 years ago. One major concern in real/virtual image blending is the need for accurate optical calibration of the endoscope-camera pair, and reliable registration onto the patient’s real world. Calibration can be tricky, needing to be redone each time the slightest modification is made to the optical system, such as rotating the camera relative the endoscope, even unintentionally.
A robotized system has the inherent advantage of automating calibration, for instance by taking snapshots of a reference grid with various positions of the endoscope-camera pair, then computing intrinsic (tangential and radial distortion values of the optics) and extrinsic (transformation equations to match the patient’s world) parameters of the whole system. Once calibrated and registered, the endoscope allows data to be overlaid and displayed on the video images, augmenting their information content ( Figure 10.5 ).
Endoscope fulcrum points and reaction forces/torques
Numerous studies of the working pivot and workspace geometries of laparoscopes have been published [ ], but very few are adapted to endonasal surgery, although these considerations are of the highest importance in devising endonasal robotic solutions. One of the most important considerations is the fulcrum point around which the endoscope rotates within the nasal tract. Eichhorn [ ] described this area as a surface with dimensions 3.93 × 2.31 mm, but reported a working volume of only 16.59 × 11.38 × 6.30 mm in functional endoscopic sinus surgery (FESS), which is much less than any conventional FESS surgical field. Further, the endoscope angle used for measurement was not reported, voiding the usefulness of this analysis.
Trévillot et al. [ ] reported much more interesting data from ex vivo cadaveric procedures, showing that the pivoting location and volume of a 30° angled endoscope depends on the anatomic site to be treated. For these authors (and referring to Figure 10.6 ), the dimensions of the bounding boxes containing the pivoting motion vary from 4.8 to 20.9 mm along the x axis, from 13.8 to 30.9 mm along y and from 7.2 to 34.9 mm along z.
Our own experimentations, performed during real surgery and recorded with a navigation system, suggest similar conclusions but with a slightly different formulation. We focused on the real surgical conditions of basic (antrostomy, radical ethmoidectomy, sphenoidotomy) and less common procedures (Draf-3, ethmoid/skull base carcinoma resection, infratemporal fossa approaches). Our set-up included a navigation system to record the positions of 30° and 45° endoscopes in the patient reference frame, with off-line computation of the geometric fulcrum minimizing the general distance d ( Figure 10.7 ) between the steadiest point along the main axis of the instrument for the whole set of recorded vectors. They were recursively compared to the first computed pivot, and a subgroup was created whenever the standard deviation between the first computed pivot and a subset of instrument vectors was more than twice the mean value of any created subset. Thus, a population of pivots was created. Twenty-one different procedures were analyzed to define which geometric structure best encapsulated these fulcrums.
The inserted endoscopes ranged from 35 to 112 mm. Depending on the surgical site targeted, the endoscope angle and the surgeon’s choice to look upward, downward or laterally, the way other instruments were disposed, and the patient’s anatomy (including septal deviation), the computed fulcrum points varied significantly, but all remained encapsulated in a pseudo-conic shell, the proximal base of which was an ellipse (long axis, 23 mm) and the distal base centered in the nostril was another ellipse (long axis, 16 mm). This pseudo-cone was 14 mm in length. Fulcrum points could be intra- or extranasal. The large scatter of the distribution of these fulcrum points reflects the variety of procedures that can be performed within a very restricted corridor.
Regarding the forces and torques involved in endoscopic transnasal surgery, Eichhorn [ ], Tingelhoff [ ], Bekeny [ ] and Joyce [ ] all reported maximum values around 7 N, while Trévillot et al. [ ], using 6-axis Force-Torque sensors (ATI Nano 43® and Schunk Mini-45®, one mounted on the endoscope, the other on a suction cannula) surprisingly recorded much higher values, reaching 38 N in the ethmoid sinus with an average value of 14 N during sinonasal tract dissection on a series of 13 specimens.
Summary of experimental work in endonasal robotics
Navigated control in functional endoscopic sinus surgery: Strauss proposed enslaving an endonasal microresector to a navigation system [ ]. A power switch was automatically opened, under software control, each time the tip of the shaver was located outside a predefined workspace, previously delineated on the patient image set. Although not yet applied in real surgery, this attempt may be seen as a first step toward robotic impeachment surgery.
In 2002, Koseki presented a non-ferromagnetic endoscope manipulator for interventional open-MRI procedures [ ]. Designing such a technology was a real challenge, but the system seems quite bulky and inappropriate for any conventional endonasal surgery.
One year later, Wurm [ ] reported the first robot dedicated to paranasal sinus surgery . Despite the appealing title of “a fully automated robot”, the system was essentially an industrial workbench robot linked to a laptop. No significant data or results were reported.
In 2011, Fischer et al. [ ] reported an interesting custom endoscope manipulator , specifically designed for endonasal surgery. Its kinematics consisted of a dual 5-bar linkage allowing 4 degrees of freedom (DOF), with an added DOF for endoscope axial translation provided by a linear motor. The system, relatively compact and easy to integrate, lacked sufficient workspace, the cranio-caudal and lateral tilts being restricted to a few degrees. Input control was limited to a couple of joysticks. Unfortunately, the device was only lab tested.
A transnasal application of a concentric-tube manipulator, developed initially at Boston University was reported by Burgner [ ]. This technology allows the inserted instrument to be given complex shapes, the number of which depends on the number of hyperelastic Nitinol tubes involved in its construction (see chapter 4.3 ). Although promising, this concept can only provide large-radius curvatures and requires a cumbersome actuating stage ( Figure 10.8 ) for moving and translating the tubes. At its present level of development, it is only used in experimental sellar surgery.