Robotics in neurosurgery

Abstract

One of the first surgical specialties to adopt robotic procedures and one that continues to innovate

Robotics is a fast-moving discipline, which – in tandem with advances in artificial intelligence and machine learning – is transforming the practice of neurosurgery. There is hope that robotics will eliminate mechanistic errors, reduce operating times and provide the same or even greater resective margins with minimal-access surgery. This will mean excellent surgical results with minimal complications. Although we can see some positive steps in this direction, there is still a long way to go, owing to the anatomical complexity of the brain and the spatial limitations inherent in neurosurgical procedures.

Interestingly, it was in the neurosurgical field that the first robotic device was used. The Unimation PUMA (Programmable Universal Machine for Assembly) 200 robot, a machine designed for industrial use, was used to position a needle precisely using CT guidance in a 52-year-old male when performing a stereotactic biopsy of a deep intracerebral lesion.¹ Although needle advancement and biopsy were done manually, the ability of a robot to position the biopsy needle precisely was a technological first. This was soon followed by the use of the same robot (PUMA 200) to act as an assistant by helping retract delicate neural structures during the surgical resection of low-grade thalamic tumours in children.² The first FDA-approved robotic device designed specifically for neurosurgical use was the NeuroMate robot (Integrated Surgical Systems, Sacramento, California, US). With regard to spinal surgery, several robotic devices exist that provide tool guidance and confirmation of implant placement.

Although there are many robotic systems currently on the market for application in neurosurgery, they can all be broadly classified into three categories: the telesurgical robot, the supervisory surgeon controlled robot, and handheld shared/controlled systems.

The telesurgical robot (master–slave)

In this type of robot, the surgeon remotely controls the actions of the robot. The NeuroArm (University of Calgary, Alberta, Canada) holds tremendous promise. It is a MRI-compatible robotic arm that mimics the movements of a surgeon’s hands. It uses piezoelectric motors and has eight degrees of freedom (DOF). This has been under steady development with purpose-designed microsurgical instruments (which are force-sensing and force-calibrated) now added to the robotic arm. Initial promising studies with rats³ have led to recent use with human subjects.⁴ It is the first robot that provides tactile feedback and is controlled by the neurosurgeon who works from a remote workstation outside the operating room. It has been reported to have been involved in more than 1,000 neurosurgical procedures, including MRI-guided tumour biopsies, microsurgical dissection and haematoma evacuations.⁵

The supervisory surgeon-controlled robot

Here the robot assists the surgeon to carry out precise tasks. After starting out with the PUMA robots in the mid-1980s, these first prototypes remain the most widely used robots in neurosurgery today. Other robots have subsequently been developed, such as the Minerva and Pathfinder robots. These largely perform stereotactic tasks with or without a frame, and have progressed from guiding biopsy needles and depth electrodes to planning and inserting pedicle screws in the spine. Supervised robots like the SpineAssist and Renaissance systems (Mazor Surgical Technologies, Caesarea, Israel) are now widely used in spinal instrumentation and, more recently, have been approved for use in the brain.

Handheld shared/ controlled systems

This is where the surgeon and the robot jointly control the instruments used to manipulate and dissect the brain structures. Therefore, the precise actions of the robot can be used in conjunction with the manipulative skills and manual dexterity of the neurosurgeon. In a way, it is akin to having the best of both worlds. There are only a few such systems in development. The Steady Hand System (John Hopkins University, Baltimore, Maryland, US) is a typical example. This instrument is held by the surgeon and the robot, and allows finer dissection and elimination of tremor and muscle fatigue. Other devices include the Evolution 1, which can be controlled to perform endoscopic procedures. The NeuRobot (Shinshu University, Matsumoto, Japan) is a remotely controlled device consisting of an endoscope, which is equipped with twin tissue forceps that can aid in tumour resection.

The ROSA^® system (Medtech Zimmer) is similar but needs the development of a whole range of jointed microinstruments and bipolar forceps before it can be used in dissection of the brain. The haptic feedback these devices provide is essential for the surgeon when microneurosurgical brain dissection is taking place.

Our experience with the ROSA^® system has shown excellent results in terms of the accuracy of electrode placement and biopsies. The planning for the surgery is done before the operation – sometimes days before. Recent imaging is imported from the internet, PACS, DVD or USB memory stick. Although there is evidence of accurate surgical plans based purely on MRI imaging, most still prefer to use image sets from fused CT and MRI scans. The plan sets out the entry point and target, and the surgeon can choose the exact path the instrument/electrode should take. There is a facility in the planning software that lights up important vascular and neural structures so they can be avoided. The plans may be exported to a portable memory drive and brought to theatre at the time of surgery.

In theatre, the plans are loaded onto the robot, and the patient’s head fixed in a Mayfield^® clamp and attached to the robot. This arrangement is unchanged for the duration of the operation. The robot registers the surface markings of the exposed areas of the face, which is similar to how standard neuronavigation systems perform registration. This process is commonly guided through an optical device in the robotic arm and there is no contact with the patient. Bone-anchored fiducials may also be used. Intraoperative imaging like 3D spin x-rays (for spinal instrumentations) can be fused during surgery to further improve accuracy.

Safety zones can be planned in, thereby preventing surgical instruments from breaching those areas and injuring vital structures

The system is capable of executing and planning the movement of surgical instruments held by the robot arm in terms of amplitude, speed (micro-and macro-adjustment) and trajectory in space, and allows the surgeon to navigate and guide the instruments held by the robot arm (freehand navigation) in a dedicated mode. This is aided by haptic feedback, which is intuitive and allows fine positioning of the robotic arm with minimal exertion.

A robotic system of this type is perhaps the commonest system used worldwide in neurosurgery. It is extremely useful for stereotactic procedures: insertion of DBS electrodes, biopsy of deep-seated complex tumours, insertion of multiple depth electrodes for stereoencephalography (SEEG), placement of microcatheters for targeted chemotherapy in gliomas, and insertion of pedicle screws in the spine. It can be used in conjunction with neuroendoscopy to help in navigating narrow corridors of access without any deviation. Safety zones can be planned in, thereby preventing surgical instruments from breaching these areas and injuring vital structures. This obviously makes the operation safer and prevents procedural complications. Many studies have been comparing the accuracy of pedicle screws and depth electrodes placed with the robot against placement by more conventional methods. They show excellent accuracy with robotic placement once the learning curve has been overcome.

Discussion

Significant advances have been made in neuroendoscopy and minimal-access surgery in the brain. Studies from Reisch,⁶ Perneczky⁷ and Kassam⁸ have shown it is possible to perform complex microneurosurgical tasks through an endoscope. This progress is due to advances in optics and the development of customised microinstruments, allowing dissection, cautery and manipulation of tissue through an endoscope tube. Adding a robot to the mix promises significant advances. At present, the robot is used mainly as a stand, freeing up a surgeon’s hand and reducing muscle fatigue. The advantage over using a conventional stand for endoscope is that the robot can mark out a ‘safety zone’, which stops the surgeon from straying and accidentally injuring important structures. Restricting movement according to defined axes or remaining within pre-defined volumes makes the surgical act safer (axial and isocentric restrictions and safety zones). This is particularly useful in neurosurgery where almost all operating is done through narrow corridors. This feature is beneficial when training juniors and will perhaps allow them the freedom to operate with less interference from their anxious trainers.

Figure 1.

ROSA^® robot

Figure 2.

Optical surface registration with the robot

Figure 3.

Surgical planning for endoscopy on the left and multiple electrode insertions for SEEG on the right

What then are the advantages brought to neurosurgery? A careful evaluation is needed so we do not disregard the tremendous advances brought by the advent of microscopy and more than 40 years of microneurosurgery. The obvious advantage is that the physical limitations of the surgeon can be neutralised. The surgeon’s tremor, tiredness and errors of depth of field will not negatively affect performance, and this will reduce complications and improve outcomes. It will be easier to execute a meticulously planned operation with little variation. The minimal access and reduced complications have seen a change in practice in epilepsy surgery, where depth electrodes inserted with a robot is replacing the use of grids and strips inserted through craniotomies. Intracranial monitoring of epilepsy with intradural grids and strips did not always yield the necessary data needed to confirm the hypothesis. However, much wider sampling is possible with multiple depth electrodes and SEEG, leading to a much higher diagnostic yield.^9,10

Precise placement of the electrodes in lesions like hypothalamic hamartomas and nodular heterotypia allows us to record from these structures, and makes it possible for ablative procedures like radiofrequency ablation or MR-guided laser-assisted ablation. The results from such procedures are encouraging. In addition, precise mapping of anatomical white matter tracts and functional networks in the brain is now possible using recordings from depth electrodes and corticortical evoked potentials.

There are some technical limitations; in endoscopy, the access channel is narrow and the permitted movements can only be coaxial. The brain does not lend itself to gas insufflation. Therefore, bimanual manipulation is difficult and manual dexterity is reduced. There is also the danger that over-reliance on the robot can reduce the situational awareness of the neurosurgeon, leading to errors that can be difficult to correct. The learning curve for such procedures is also steep.

Neurosurgery is a constantly evolving discipline and perhaps the first to innovate and embrace new advances. Exciting advances in Augmented Reality and Artificial Intelligence will no doubt come together in an integrated platform to make the technical act of surgery more precise and free from errors. The neurosurgeon of the future will be more a neuroscientist, collecting and interpreting vast amounts of information sampled from multiple electrodes, mapping out networks of the brain, and helping us to gain a better understanding of how the brain works.