Which operation in neurosurgery is most likely to be robotically automated, and why?
Written By June Hyun
Robotics has a long and progressive history intertwined with neurosurgery. In 1985, the first surgical robot, PUMA 560, performed a brain biopsy before transitioning to prostate-related treatments - the most commonly performed robotic surgery. The UK performed over 49,000 robot-assisted surgeries in 2022, and numbers are only rising.
However, the current gold standards, such as the Da Vinci system, have low levels of automation - most are in a "master-slave" relationship where the robot is under the surgeon's control. After the advent of accessible AI, surgical researchers have begun clinical trials attempting to increase automation through existing toolkits. AI developed for surgical use magnifies, translates camera views, and highlights surgical margins without input (known as Level 2 Partial Automation). Trials at Children's National Hospital in Washington, D.C. have successfully automated bowel anastomosis (Level 4 High Automation) in porcine models with a post-operative prognosis that is similar to or better than that of laparoscopy. Automation could improve efficiency in theatres and free surgeons for higher-ability tasks while minimising time spent on high-volume, rule-based and repetitive procedures. However, to further develop automated surgical techniques, data must be structured and digitised in a simple process (e.g., visual input - binary conversion pathway) within reach of technologies that are currently being developed. Routine, minimally invasive neurosurgical procedures - such as Deep Brain Stimulation (DBS) and Laser Interstitial Thermal Ablation (LITT) - are potential candidates for eventual Level 5 Complete Automation, improving patient outcomes by reducing operation times and surgeon error.
DBS is a standard surgical procedure for medication-resistant epilepsy, OCD and essential tremor. Artificial electric currents from electrodes stimulate neurones in the brain, leading to therapeutic effects through mechanisms that are not yet fully understood. DBS is carried out via two to three separate operations. The surgeon implants electrodes into pre-selected regions of the brain while threading the leads subcutaneously to the posterior skull. A second procedure connects the threads to a pacemaker implanted into a pouch below the clavicle. Existing robotic platforms, such as ROSA ONE (a leading neurosurgical platform assisting surgeons in pre-operative planning and minimally invasive procedures), provide reliable and well-tested robot-assisted surgical approaches. The software identifies ideal lead placements via CT and MRI tomographies and serves as a stable rod for the surgeon to guide into the brain, reducing operation times by three hours. The high reliance on imaging, relatively low complexity (as the operation is confined to a very localised brain region), and need for repeat operations (electrodes require recurring surgeries to replace batteries) suggests that DBS could be a strong candidate for full automation within the next decade, from trajectory planning and electrode placement to anastomosis. Indeed, minimally invasive procedures can enable single-day discharges, enhancing healthcare quality for patients and reducing inpatient stays. Automation could also make neurosurgeons more available for more complex, less routine procedures, allowing for better allocation of highly skilled surgeons in a stretched NHS. Laser Interstitial Thermal Ablation (LITT) is of great utility for catheter placements and is a core weapon against deep-seated lesions and tumours. Small probes use laser-generated thermal energy to ablate brain tumours or specific brain regions (e.g., for epilepsy treatment), reducing the need for craniotomies and lowering the risk of infection. Current platforms such as Stealth Autoguide (Medtronic) provide stereotactic guidance. The surgeon is responsible for managing the probes and margins, determining how far to push the boundary based on MRI and visual data. The need for consistency, stability and free range of movement in confined spaces highlights the potential for robotic assistance and automation. Indeed, groups at the Miller School of Medicine demonstrated ROSA ONE’s promising performance in non-framed stereotactic placement of laser catheters in forty-three patients, reducing post-operative complication risks. Furthermore, existing and developing AI margin-recognition technologies suggest that, soon, robots could be autonomously deciding regions of ablation under supervision. LITT automation would decrease operation times, push the boundaries of tumour resection and improve the overall prognosis for patients undergoing oncological operations.
DBS and LITT both heavily rely on stereotactic data. The mass of digitised data such as the EHR and existing robotic platforms (e.g., ROSA ONE) suggest that the two procedures have an equal chance of automation. However, current tumour margin delineation techniques limit the advancement of LITT. Fluorescent markers such as 5-ALA have reported false positives in normal tissue. It is still difficult for surgeons to characterise presenting tissue types, even with intraoperative imaging and prior MRI and PET scans. Since AI learns from existing data, current challenges in demarcation suggest that LITT automation would require significant improvements in intraoperative tissue characterisation. An alternative, such as biopsies for concrete delineation, would prove a challenge to digitalise and increase operation times. Thus, DBS's relative simplicity, repeatability, and high turnover rate (having a short operation-to-discharge time) makes it the most suitable procedure for automation in the near future.
Read the rest of the edition here
