Keywords
endoscopic neurosurgery - pituitary surgery - robotic surgery - skull base
Introduction
Endoscopic endonasal surgery (EES) offers several advantages over traditional transcranial
approaches when treating skull base pathologies, including reduced exposure and neural
tissue manipulation.[1]
[2]
[3] Indeed, the endoscopic transsphenoidal approach (eTSA) has become an increasingly
common approach to the sella, and the expanded endoscopic endonasal approach (EEEA)
has extended the surgical reach beyond the sellar region to include the anterior cranial
fossa, parasellar region, clival region, and craniovertebral junction.[1]
[2]
[3] Nevertheless, EES is challenging due to the anatomical constraints enforced by the
long, narrow nasal corridors, which impose a fulcrum effect on surgical instruments.[4] Additionally, current surgical instruments have a limited ability for tissue manipulation
due to their lack of articulation.[5] Consequently, EES has a steep learning curve.[5]
To overcome these challenges, we have developed a handheld robot consisting of a miniaturized,
articulated end-effector coupled with an ergonomic controller designed to offer an
expanded workspace and greater dexterity in endoscopic neurosurgery. Previously, the
device was found to improve maneuverability, was deemed easy and comfortable to use,
and was preferred over conventional endoscopic instruments when evaluated by neurosurgeons
with prior experience with EES.[6]
However, certain limitations were identified, including insufficient force delivery
during robotic joint actuation, challenges with precision, limited diagonal workspace
reachability, and tissue entrapment within the joints. Additionally, the joystick
was deemed to lack appropriate sensitivity.[6] Therefore, the robot has undergone significant design and manufacturing updates,
including larger motors, end effector re-design, and more sophisticated control software.
In this study, we present an evaluation of the updated handheld robot (version 0.2).
First, we perform a preclinical randomized cross-over study to compare the robot against
standard endoscopic instruments in a phantom pituitary tumor resection task. Second,
we conduct an iterated cadaveric study to evaluate surgeon perceptions of the engineering
updates.
Methods
Device Development
A detailed description of the development and re-design of the robot has been published
separately.[6]
[7] In brief, the handheld robotic system is composed of an ergonomically designed,
handheld controller with a rotating joystick body that can be placed at the position
most comfortable for the user, its accompanying control box, and articulated 2 or
3 degrees-of-freedom end-effectors utilizing a wide range of interchangeable articulated
instruments at its distal end. For this study, only the curette end-effector was used,
as the device was compared with a standard curette instrument. Since the previous
study, key updates include integrating larger motors, an end-effector re-design to
allow for a wider effective workspace, incorporation of closed-loop control both in
position and velocity modes, parameter tuning, and return-to-neutral functionality.
The handheld device is shown in [Fig. 1].
Fig. 1 Depicts the handheld robot with its articulated distal end, interchangeable end-effector,
trigger and rotating-joystick body.
Participant Recruitment
Participants were recruited from a single university teaching hospital and were defined
as (1) experts if they had completed neurosurgical training, (2) intermediates if
they were neurosurgical registrars (residents) with experience in pituitary surgery,
and (3) novices if they did not meet either of the previous criteria. Demographic
data, including the stage of training, sex, and handedness, were recorded.
Phantom-Based Comparative Study
Study Design
The study was approved by the local ethics committee (reference: 17819/001). The phantom
study compared the robot against standard endoscopic instruments in a pituitary tumor
resection task. Eight participants were recruited. Due to pragmatic constraints in
trial design, no formal power calculation was performed. A randomized cross-over study
was conducted using a validated phantom model for the eTSA (UpSurgeOn Transsphenoidal
[TNS] Box).[8] The TNS Box was prepared by creating a surgical corridor such that the tumor was
accessible. Participants were tasked with performing a tumor resection after a short
standardized introduction to the robot and standard instrument.[9] This was performed using either the standard instrument (a bayonet-shaped, 45° angled
ring curette, model: FA061R, BBraun, Germany) or the robot. Participants were given
3 minutes for tumor resection. The end-effector for the robot consisted of a ring
curette with similar dimensions to the standard instrument. Participants performed
the resection task five times using each instrument. The tumor and dura were replaced
for each iteration, but the rest of the model remained the same. Tumors were weighed
before and after the resection task to calculate the extent of resection (EOR).
Following the resection tasks, participants completed a validated, surgery-specific
task load index (SURG-TLX[10]), which prompted them to consider the extent to which they experienced various intraoperative
workload domains. Workload domains included mental, physical and temporal demands,
task complexity, situational stress, and distractions. Participants completed weighted
ratings of each subdomain, which were aggregated to produce a total workload score.[10]
Finally, participants completed a post-task questionnaire to score each instrument
regarding ease of use, comfort, precision, and integrity. The scores ranged from 1
to 5 points (1 = poor, 5 = excellent). The overall preferred instrument was also assessed.
Outcomes
The primary outcome was the median EOR achieved on the final (fifth) attempt. The
difference between the first and final attempt was used as the comparative metric
to account for potential learning curve effects. Secondary outcomes included the median
EOR on the first attempt, the composite SURG-TLX workload scores, and post-task questionnaire
outcomes.
Statistical Analysis
R Studio (2022.07.2) and Excel (Microsoft, version 16.6.1) were used for data analysis
and data representation. Outcome measures, including the EOR and SURG-TLX, were expressed
as median (interquartile range [IQR]) and statistical differences were assessed using
nonparametric tests (paired, Mann–Whitney U-test). For parametric data distributions
(e.g., post-task questionnaire), paired t-tests were performed. Categorical variables, such as the favored instrument, were
evaluated using a chi-squared test. p <0.05 was deemed statistically significant.
Cadaver-Based User Study
Cadavers were used according to the local anatomical board's ethical requirements.
The objective of the cadaver-based study was to qualitatively evaluate the feasibility
and acceptability of the device through the exploration of surgeon opinions regarding
domains pertaining to the robot's design updates. One fresh and frozen cadaver was
obtained for the study, serving as the subject for all participants in the surgical
tasks. The eTSA, including the durotomy, was performed pre-task by the expert neurosurgeon.[9] Following this, participants explored various functional domains of the device by
completing simple tasks, detailed in [Table 1]. Participants performed each task using two modes of end-effector control: position
mode and velocity mode. In position mode, the end-effector follows the position of
the joystick, meaning when the joystick is released, the end-effector is at a neutral
position. In velocity mode, the joystick determines the desired change in position
of the end-effector, meaning when the joystick is released, the end-effector maintains
its position. Participants were interviewed while performing each task to gather their
opinions with respect to the device's: (1) reachable workspace, (2) precision, (3)
force-delivery, (4) ease-of-use, (5) structural integrity, (6) drawbacks, and (7)
benefits compared to standard instruments. Their answers underwent a reflexive, thematic
analysis.[11]
[12] All interview questions can be found in [Supplementary Table S1].
Table 1
Cadaveric tasks
|
Task
|
|
Using the joystick, explore the vertical, horizontal, and diagonal movement capabilities
of the robotic end effector.
|
|
On the video display of the endoscope, the assessor will now ask you to touch 4 points
inside the sphenoid sinus.
|
|
Place the end effector against the mucosa of the sphenoid sinus. Then, using the joystick,
assess the applicable force at the hinged end-effector moving the sphenoid sinus mucosa
and pituitary gland.
|
|
The surgical tasks are now complete, you may interact with the tissue environment
as you wish.
|
Results
Phantom Study
Demographics
Participants included six males and two females. There were two experts, two intermediate,
and four novice participants, of which seven were right-handed and one was left-handed.
Extent of Resection
On the first attempt, participants using the standard instrument and robot achieved
a median EOR of 84% (IQR: 65–91%) and 58.6% (IQR: 52–77%), respectively (p = 0.055). On the final attempt, the median EOR with the standard instrument and robot
were 80% (IQR: 70–89%) and 83% (IQR: 61–94%), respectively (p = 0.76). The EOR improved across attempts with the robot (p = 0.38) and decreased with the standard instrument (p = 0.95) ([Fig. 2]).
Fig. 2 Graph displaying the extent of phantom tumor resection achieved with the standard
endoscopic instrument versus the handheld robot. The black circles signify median
values for the respective instruments.
Experts and intermediates outperformed novices on the first resection attempt using
both the standard instrument (87% vs. 76%, p = 0.25) and the robot (71% vs. 55% p = 0.25). Novices achieved a first-attempt EOR of 76% (IQR: 60–86%) with the standard
instrument and 55% (IQR: 52–91%) with the robot (p = 0.62) ([Fig. 3]). On their final attempt, novices resected 80% (IQR: 70–84%) with the standard instrument
and 83% (IQR: 64–91%) with the robot (p = 1.0). Intermediate and expert surgeons achieved a greater first attempt EOR with
the standard instrument (87%, IQR: 77–92%) compared to the robot (71%, IQR: 56–85%,
p = 0.12). On the last attempt, intermediates and experts achieved an EOR of 80% (IQR:
70–91%) with the standard instrument and 80% (IQR: 61–94%) with the robot (p = 0.88).
Fig. 3 Graph displaying the extent of phantom tumor resection achieved with the standard
instrument versus the robot. Grouped by level of experience. The black circles signify
median values for the respective instruments.
SURG-TLX Outcomes
Compared to the standard instrument, the robot was associated with a lower mean total
workload score (standard =156, robot =118, p = 0.006). Considering subdomains, the robot was associated with a statistically significant
reduction in perceived physical demands (robot = 5.4, standard instrument = 42, p = 0.03). Otherwise, subdomain differences were insignificant.
Post-task Questionnaire
[Fig. 4] depicts the questionnaire results. All but one participant (7/8) favored the robot
over the standard instrument overall (p = 0.0027). The reported robotic benefits included that it had a “better range of
movement,” “required fewer wrist movements,” “was more comfortable,” and “more precise.”
Reported drawbacks of the robot included that it was “uncomfortable to hold the thumb
on the joystick,” “(it) clashed with the endoscope,” “would benefit from a longer
curette,” and required “haptic feedback.”
Fig. 4 Bar chart displaying the user ratings of instruments across four domains.
Cadaveric Study
Demographics
Participants included one expert, one intermediate, and four novices. Two participants
were female, and four were male. Five were right-handed, and one was left-handed.
Qualitative Outcomes
[Fig. 5] depicts the robotic manipulation of the cadaveric tissue. All surgeons (n = 6/6) found the range of movement of the robot sufficient in all directional planes,
including the diagonal movements, which were deemed to be insufficient in the previous
device prototype.[6] One expert participant highlighted that “this feature could reduce the amount of
instrument changes performed during an operation.” All surgeons found the end-effector
capable of applying sufficient forces at a predefined angle and during movement. Specifically,
the expert neurosurgeon commented that the device was “more than capable” while another
participant highlighted the device was “great for soft tissue manipulation” and that
it could “even be used to scrape bone.” Additionally, all participants found the instrument
easy to use. Finally, when asked to pick a preferred end-effector control mode (i.e.,
position or velocity mode), participants found the modes equally capable across domains,
including articulation, ease of use, precision, and force delivery.
Fig. 5 The articulated end-effector inside the cadaveric specimen (a) actuated in different angles to test reachability and (b) interacting with soft tissue to test force-delivery.
Limitations of the robot were also highlighted by participants. First, the device
handle was observed to occasionally clash with the endoscope, disturbing the operative
workflow. Also, the lack of end-effector variety was highlighted. Additionally, the
end-effector showcased minor structural damage during the last attempt by the last
participant, suggesting the need for reinforcements. Finally, careful examination
of the end-effector actuation suggests the presence of mechanical backlash.
Discussion
Principal Findings
In this study, we present an updated prototype (version 0.2) for the first-of-its-kind
handheld robot designed for endoscopic neurosurgery. We compared the robot against
standard instruments through a high-fidelity phantom study and established clinician
perspectives of the engineering updates through a cadaver-based user study. The robot
showed noninferior effectiveness and superior perceived workload whilst offering other
potential advantages. Additionally, we identified domains for further iterative developments
ahead of creating a prototype ready for a first-in-human assessment.
In the phantom study, we observed a learning curve effect, as the robot was inferior
to the standard instrument on the first attempt at tumor resection (84% vs. 59%, p =0.055) but equivalent on the fifth attempt (80% vs. 83%, p = 0.76). These learning curves will be used to guide clinician training with the
robot prior to future clinical trials. Separately, participants scored the robot higher
in terms of range of movement, ergonomic manipulation, comfort, ease of use, and precision.
Indeed, the robot was associated with a lower total cognitive workload (p = 0.003), which may explain why users demonstrated a significant improvement with
successive resection attempts using the robot, unlike with the standard instrument,
with which resection outcomes decreased slightly. Overall, the robot was the favored
instrument by 7/8 participants (p = 0.0021), implying clinician usability, a critical factor in the evaluation of early-stage
medical devices, as outlined by the IDEAL-D framework.[13]
In the cadaver-based user study, participants expressed that in the context of the
eTSA, the device achieved a sufficient range of movement in all directional planes
and applied an appropriate amount of force with its end effector. Additionally, the
expert surgeon highlighted that due to the expanded robotic workspace, one could envision
that the device would reduce the number of instruments used intra-operatively. Such
effects may improve operative workflow efficiency and thus reduce operative times.
Overall, despite the need for further iterative developments, the device demonstrated
clinical feasibility and user acceptability.
There were also drawbacks identified with the current device prototype, which are
to be addressed in future iterations. The device's end effector experienced minor
structural damage during the cadaveric study, indicating a need for a reinforced design.
Furthermore, the larger size of the robotic handle resulted in clashes with the endoscope,
which may limit the device's applicability to mono-nostril approaches.
Comparison with the Literature
Despite significant advances made in robotic-assisted surgery, the translation of
such systems into endonasal skull-base surgery has been slow.[14] While studies have demonstrated the feasibility of robotic endoscope holders,[14] the only robot capable of tissue manipulation that has undergone clinical trials
for skull-base surgery is the DaVinci system (Intuitive, Surgical).[15]
[16] However, the Da Vinci was built for general surgery and thus faces challenges related
to its instruments' size, operating room footprint, and lack of dedicated tools (e.g.,
drills) when deployed for pituitary surgery.[15]
[16] Indeed, the DaVinci cannot be inserted through the nose and thus must currently
rely on transoral approaches to the sella.[14]
Examples of robots purpose-built for endoscopic skull base surgery can be found in
the preclinical stage of translation, in which continuum robots are popular. Such
teleoperated “flexible robots” consist of thin, flexible, tubular shafts with interchangeable
end-effectors capable of tissue manipulation.[17] Continuum robotic shafts may consist of pre-curved, concentric, tubular segments
(controlled through the telescoped elongation and axial rotation)[18]
[19] or nitinol tubes that can be flexibly adjusted.[20] To date, preclinical studies have validated some of these designs in phantom tumor
resection tasks conducted by a single expert.[18] However, continuum robots present their own issues related to sterilizability, controllability,
and the need for a support base, which may limit distal-end dexterity and force delivery.[17]
[21] Handheld robots bypass many of these challenges.
In our study, we present the first-of-its-kind, preclinical, handheld robot for endoscopic
skull base surgery, which offers several advantages over the aforementioned teleoperated
alternatives. First, handheld robots do not require a support structure or base, reducing
surgical workflow disruption. This “easy” integration into the operative workflow
may have a profound effect on the human factors associated with robotic-assisted surgery,
as avoiding the physical separation of the surgeon from the surgical team (as in tele-operated
systems) reduces the demand for verbal communication, cited as the most common cause
of procedural error and surgical injury.[22]
[23]
[24]
[25]
[26] Second, the robot has a smaller footprint, which can reduce purchasing and maintenance
costs. Finally, it has a small and lightweight design, resembling typical surgical
instruments and thus can more easily be adopted in procedures where frequent tool
changes are required, such as the EEEA.[27]
Since the first study describing our handheld device, engineering updates have addressed
previous limitations related to diagonal workspace reachability and effector force-delivery.
As demonstrated in our cadaver-based surgeon interviews and phantom resection tasks,
our device is easy to use and clinically feasible. Indeed, surgeons across the continuum
of surgical experience preferred the handheld robot over conventional tools.
Strengths and Limitations
Our study's strengths relate to the surgically relevant context in which the engineering
updates were evaluated. In the phantom study, experts outperformed novices at the
surgical resection task, implying the TNS model was reliably able to distinguish surgical
performance (whether due to skill or instrument differences). Also, the cadaveric
study enabled users to explore the device's functional domains in the highest fidelity
setting possible. This suggests the benefits observed in our study will be replicated
in future in-human evaluations.
Our study also has limitations. First, the pragmatic constraints in the number of
recruited participants limited our results' statistical significance. Additionally,
the number of resection attempts in the phantom study was low; hence it is unknown
if the improved resection performance observed with the robot would have continued
with further practice. Separately, due to the lack of realism in the resection task,
some participants used unrealistic gestures, such as squeezing the tumor out of the
synthetic dura. Finally, participants were not blinded to the trial's intent, which
may have biased subjective interviews.
Conclusion
This study presents an updated prototype for the first-of-its-kind handheld robot
designed for endoscopic neurosurgery. The device demonstrates its clinical feasibility
and user-acceptability when compared to standard endoscopic instruments. Additionally,
interviewed neurosurgeons reported sufficient workspace reachability, force-delivering
capabilities, and precision. Overall, we demonstrate that the device is comfortable,
easy to use, precise and clinically effective, yet has scope for further iterative
improvements.
