Assessing visualization in robotic-assisted surgery: demystifying a misty lens

Despite major technological advancements in robotic-assisted laparoscopic surgery (RAS), there remain shortcomings yet to be addressed. This study assesses the prevalence of suboptimal vision in minimally invasive RAS and corresponding factors regarding related surgical conditions. 45 minimally invasive robotic surgeries, performed using Da Vinci XI, were observed across three surgical subspecialties: general, urology, and OB/GYN. Lens occlusion events were monitored and defined as the presence of a visual distortion caused by debris deposition on the scope lens. Lens occlusions and cleanings, and “active instrumentation” were recorded. Descriptive statistics summarized duration-based variables, and one-factor ANOVA compared the presence of active instrumentation. Cases averaged 127 ± 76 min. Active instrumentation ANOVA during lens occlusions demonstrated significant variation between categories (F7, 256 = 11.63, p = 2.558e-13). Post hoc Tukey HSD found electrocautery devices were active significantly more during occlusion events (37.9%) than other instruments. On average, lens cleaning occurred every 36.5 ± 39.8 min despite lens occlusion occurring every 24.5 ± 15.7 min. Of the operative time observed, 41.4% ± 28.1% was conducted with visual distortion. 1.16% ± 0.97% of time observed was spent cleaning. Although only 1.16% of operative time was spent cleaning, surgeons experienced suboptimal conditions for nearly 35× the time it would take the clear lens, potentially indicating a tendency to avoid cleaning the lens to disrupt surgery. Future research may examine the impact of occluded visualization and lens cleaning on other aspects of surgery.


Introduction
Robotic-assisted laparoscopic surgery (RAS) is a rapidly growing field within the surgical landscape due to its efficacy, affirmed by continued evidence demonstrating improved patient outcomes [1][2][3]. Hospitals have been quick to adopt RAS systems given the benefit provided to patients. A cohort study of 73 hospitals involving over 150,000 patients showed an increase in RAS from 1.8 to 15.1% of all general surgeries performed between 2012 and 2018 [4]. Furthermore, upwards of 693,000 robotically assisted surgeries were performed in the US alone in 2017 [5]. Current projections suggest that RAS usage will further expand between 2020 and 2026 with a compound annual growth rate of 22.18%; this is reflected in its market growth from a 2020 valuation of USD 8.307 billion to a projected USD 28.34 billion in 2026 [6]. Worldwide, the most popular robotic surgery platform is Intuitive Surgical's Da Vinci robot, although both Johnson & Johnson [7] and Medtronic [8] have recently entered the space.
The rapid adoption of the Da Vinci robot in RAS systems can at least be partly attributed to some key patient and surgeon benefits-including greater precision, shorter recovery times, and reduced pain and bleeding-which are linked to the technological features of the robot [9]. One primary area of innovation is vision enhancement. During surgery, a display monitor exhibits a 3-dimensional high-definition video feed of the surgical area through the hood of the surgeon console; the camera position is controlled via a toggle on the hand control activated by an available foot clutch. Despite the efforts that have yielded such improvements in visualization, there are still some shortcomings in this arena [1]. One of the most common and frustrating issues is visual impairment requiring pausing to clean the scope. Suboptimal visualization as a result of lens occlusion has been explored most extensively for the case of traditional, or Straight Stick (SS), laparoscopic surgery where common mechanisms for lens occlusion include condensation, debris splatter or buildup, and bodily fluid deposition on the lens [10]. Visualization challenges in RAS have key similarities and differences from those faced in SS laparoscopy. Although both depend on camera visualization of the surgical site, the RAS camera is robotically controlled directly by the surgeon and differs in aspects such as scope optics, geometry, and some general functionality/add-ons. Additionally, the method for lens cleaning differs in RAS relative to SS laparoscopy in that the camera interfaces with a robotic arm in addition to the body port. This adds additional steps to the cleaning process such as unclipping the camera from the arm and ensuring that, when replacing the camera, it interfaces correctly with the robotic arm. Despite the notable variation observed between SS and RAS, there is a dearth of materials specifically assessing RAS visualization in the present literature.
Scope lens cleaning has been encountered as a point of difficulty in both SS and RAS laparoscopy. While the literature has mainly focused on this issue in SS laparoscopy, its prevalence in RAS warrants further exploration as well. This is supported by studies examining the impact of poor visualization on surgical safety, focusing on potential complications during the procedure as a result of lens occlusion and subsequent cleaning. Sutton et al. showed that surgeons experience a high rate of gaze disruptions from their central task owing in part to obscured vision and subsequent instrument cleaning/exchange [11]. Furthermore, an observational study by Yong et al. revealed that 56% of laparoscopic surgery is spent without a clear display on the surgical monitor [13]. The impact of this issue is clearly felt by surgeons who report feeling frustrated due to repetitive cleaning, an action that doubly results in an interruption in surgical flow [11,13]. In a high-stakes environment such as surgery, the effects of poor visualization have the potential to cause harm to patients.
Given the impact of poor visualization resulting from lens occlusions, there currently exists a quickly expanding field within the medical community focused on investigating methods of lens cleaning. Because lens fogging is one of the most common issues during laparoscopy, a multitude of antifogging methods/products are employed by surgeons today, including surfactants, mechanical and material technology, and warmers [12]. For general occlusions, methods involving dry/wet cloths, warm water, or bringing the scope into contact with nearby viscera are utilized [13]. While there is extensive research on lens cleaning for SS laparoscopy, less is known about this topic with regards to the robotic space.
To add to the current body of knowledge on RAS, we designed the present study to provide a broad assessment of the prevalence of and factors contributing to suboptimal vision in RAS. Our primary objectives within this schema were to (1) explore active correlations of instrument activity and surgical specialty relative to image occlusions in RAS and (2) track debris events to observe associated impaired vision and lens cleaning times. Ultimately, we aimed to bridge the gap in knowledge of impaired visualization in RAS by quantifying cleaning time and occurrence of lens occlusion by various mechanisms specific to RAS.

Materials and methods
This was an observational study of 45 robotic-assisted laparoscopic surgeries split among three surgical specialties: general surgery (n = 19), urology (n = 17), and obstetrics & gynecology (OB/GYN; n = 9). All surgeries were minimally invasive and performed with the Da Vinci XI surgical robot. To investigate the surgical conditions associated with suboptimal vision in RAS, we collected data on length of time, type of active instrumentation, and surgical procedure associated with each occlusion and cleaning event throughout each RAS case. During the cases observed, the lens was cleaned using either a Da Vinci cloth or Clearify device [14].
All data were collected by tracking data on a custom survey sheet that indicated the occurrence of events during a surgical case. This included case events, lens occlusion events, and lens cleaning events. Timestamp data were simultaneously captured to aid in visualizing the frequency and duration of these events. Case events were defined as either camera insertion or camera removal, which were used in conjunction with timestamps to begin or terminate periods of data collection for the surgical case-unless the insertion/removal was related to a lens cleaning event. Lens occlusion events occurred when there was a visible distortion through the scope. Occlusions could be a result of fluid, debris, fog, or other substances accumulating on the lens. To determine if any correlations existed, the instruments active during the occlusion event were recorded. These instrument options included cautery, suction, irrigation, scissors/cutting, camera movement, instrument movement, inserting/removing an instrument, or other. The last type of event recorded was a lens cleaning event, which occurred when the scope was removed from the body to be cleaned. Lens cleanings were recorded with the method of cleaning (by Da Vinci cloth or by Clearify) and the duration of the cleaning in seconds. The two observers established criteria, as per the Definitions and Classifications section below to ensure robust alignment and standardization for all collected data. This effort mitigated the potential of inter-observer variability. This study was IRB exempt as no patient data was captured.

Definitions and classifications
We defined suboptimal vision as any moment in which the screen was obscured by a visual occlusion; relatedly, clear vision indicated no presence of such visual obfuscation. An occlusion event was defined as any additional distortion present on the screen as compared to the previous image. A cleaning event was defined as the removal of the scope from the body cavity for the purpose of clearing debris from the lens. Camera insertion was defined by the scope being inserted through the trocar to the body cavity and then attached to the robotic arm. This was recorded at the moment when the surgeon had control of the scope, which was indicated by a sound emitted by the robot. Camera removal was defined similarly, where the timestamp was recorded at the moment the camera was undocked from the robot and subsequently removed through the trocar. Aligned with the above insertion/removal definitions, the bounds of the recorded cleaning time were defined by when the scope was undocked and docked to the robot arm. An exception to this procedure was that some cases required a period of manual laparoscope driving prior to docking the robot. These periods were noted, and events were recorded in the same manner but with camera insertion time defined by when the field of view was entirely inside the body cavity. When recording cleanings, occasionally the scope would have to be cleaned multiple times because there was still debris remaining on the lens. If the scope did not meet the criteria for being fully inserted when it was removed again to be cleaned (i.e., it was not entirely inside the body or attached to the robot), the entire process was recorded as a single cleaning event. If the scope was fully attached in between cleanings, two separate cleaning events were recorded, even if one immediately followed the other.

Data analysis
A timestamp was recorded for every data submission, allowing us to quantify numerical variables such as the total duration of the case, time spent with suboptimal vision, and the frequencies of occlusion and cleaning events. Accompanying these events were additional categorical variables including the instruments that were active during occlusions and the method of cleaning. A one-way ANOVA statistical test followed by Tukey-Kramer post hoc analysis was used to test for significant differences between the myriad of factors, such as the frequency of different instruments active during occlusions and the time spent with suboptimal vision for different surgical specialties.

Results
The total operative time tracked over 45 cases amounted to 5696 min with the average surgical case lasting 127 ± 76 min. It should be noted that case length was highly variable between specialties. Urology had the highest average case length at 181 min while General Surgery and OB/ GYN, respectively, averaged 89.4 and 102 min. The average time for a single scope cleaning over all specialties was 28.7 ± 21.0 s with 156 cleaning events across all cases. There was an average of 1 cleaning event every 36.5 ± 39.8 min and 3.5 ± 3.7 cleaning events per case. Only 1.2 ± 1.0% of surgery time was spent cleaning the scope while 41.4 ± 28.1% of the operative time was spent under suboptimal vision. The percentage of operative time with suboptimal vision was highly variable between specialties as well, with Urology, General Surgery, and OB/GYN at 47.6 ± 20.4%, 42.9 ± 33.8%, and 28.4 ± 25.5%, respectively (see Table 1).
A total of 338 occlusion events across all cases yielded an average of 1 occlusion every 24.5 min of surgery and 6.8 occlusions per case. Of these, the use of cautery (37.9%), camera movement (22.2%), instrument movement (12.4%), and insert/remove instrument (9.2%) were the most frequent activities during an occlusion event ( Table 2 Column D). A Pareto Analysis (Fig. 1) affirmed that those four instrumentation activities accounted for 82% of all occlusions. None of the remaining four activities constituted more than 8% of the total occlusion count. A subset analysis included only occlusion events directly preceding cleaning events. Active instrumentation in this subset of data was compared to that among the overarching dataset including all occlusion events. Both camera movement (22.2% versus 27.2%) and inserting/removing instruments (9.2% versus 14.3%) were more likely in this subset versus all occlusion events, while cautery usage was less (37.9% versus 29.9%). Other types of activity had a less severe decrease, such as instrument movement which dropped less than 1% (12.4% versus 11.6%). A more thorough analysis of this metric for each instrument was performed by finding the ratio of the subset of occlusion events recorded prior to a cleaning event over the total number of occlusion events for the given instrument (see Table 2 column E). This is expressed in the below equation.
There was a statistically significant difference between the use of electrocautery devices and all other instrument categories during occlusion events (p < 0.05). Additionally, there was a significant difference in occlusion counts between camera movement and non-specified laparoscopic instruments (p = 0.010) and a trend toward camera movement and irrigation use (p = 0.177). All other comparisons between active instrumentation categories did not yield statistically significant results (p > 0.05). Additionally, there was a statistically significant difference between % Occlusions preceding cleaning per instrument = Occlusion count preceding cleaning Total occlusion count . occlusion counts for general surgery and urology cases (p = 0.0021) along with urology and OB/GYN cases (p = 0.0326). While occlusion counts were significantly different between specialties, the rate of occlusions per minute was not. There was no statistically significant difference between surgical specialties with regards to the time spent cleaning or the number of cleaning events. However, there was a statistically significant difference between average occlusion times for urology and OB/GYN cases (p = 0.002) along with urology and general surgery cases (p < 0.001).

Discussion
Upon assessment of the data collected, a common trend was the presence of much variability in visual acuity from surgery to surgery, supported by the standard deviations of average cleaning and occlusion times. A variety of factors may contribute to this variability including the nature of the occlusion, experience of the surgeon and robot technician, differences in patient characteristics, or operating room (OR) equipment and team structure. Anecdotal data collected during OR observations suggests that surgeon and surgical technician's comfort with RAS techniques play a major role in clear visualization. Despite not collecting quantifiable data on this metric, raw feedback suggested that some surgical technicians were able to interface with the robotic system more effectively than others; this variability may be partially reflected in the high standard deviation observed in cleaning times (28.7 ± 21.0 s). Variability in occlusion and cleaning times may also be explained by the nature and severity of the occlusion, given there are numerous types of debris variables that can lead to lens occlusion such as blood smearing, miscellaneous fluids contacting the lens, lens location for obstruction, or even manufacturing and equipment performance variations. In addition, relative standard deviations (RSDs) for time spent with occluded vision were larger for surgeries lasting under one hour (RSD = 115%) than for those over one hour (RSD ~ 50%). This may be due to shorter case times lending themselves more readily to high variation, though such a large decrease may also indicate that overall surgery length affects the variability of time spent with suboptimal vision. When assessing the potential impact of surgical specialty on laparoscopic outcomes, the data showed a significant increase in occlusion count and time spent with occluded vision for urology cases versus general surgery or OB/GYN cases. There did not appear to be a significant difference between OB/GYN and general surgery cases, though the raw data showed that OB/GYN had a longer average occlusion time. In addition, the rate of occlusions per minute was not significantly different between specialties. Given that the average length of urology cases was longer than the other surgical specialties and the occlusion rate was similar between specialties, the significantly higher occlusion counts and time spent with occluded vision in urology is likely attributed to longer case length, not a higher frequency of occlusions occurring. Despite this, an interesting dichotomy can be seen between OB/GYN and general surgery cases where OB/GYN had a longer average surgery length but a lower average occlusion time than that of general surgery. In addition to this, OB/GYN cases had a significantly lower occlusion count than urology cases. While this could be for a number of clinically related reasons, it is interesting to note that the data shows debris events occurring less in OB/ GYN compared to the other specialties observed. This difference may also be attributable to the type of case, step in the procedure, and location/anatomy of the operative area (i.e., deep within the pelvis behind pubic bone versus anteriorly). More case counts would help further explore and validate the significance of this potential trend. In contrast to occlusion counts and time, there was no observed difference in cleaning count and cleaning time between different surgical specialties. This suggests that surgeons across specialties may only be willing to clean at certain points in surgery and/or only every so often despite significant differences in occlusion counts.
In the comparative analyses performed, there was a significantly greater presence of active electrocautery devices in the surgical field during lens occlusion events than any other instrument category. The statistical significance in this dataset shows an interesting relationship between cautery activity and distorted surgical vision. Despite this putative connection, this study has raised the question of whether cautery use can be directly correlated with visualization that is poor enough to warrant subsequent cleaning. Subset analysis was performed on the data that exclusively comprised the final occlusion event before a lens cleaning. When comparing active instrumentation in all occlusion events to the subset of occlusion events, camera movement increased in proportion while cautery decreased. Further data collection surrounding the metric of occlusions preceding cleanings could validate the significance of that trend. Although active cautery occlusions made up a significantly larger proportion of total occlusion events, camera movement occlusions preceded a cleaning slightly more often. In addition, of all camera movement occlusions, 53.3% preceded a cleaning, while only 34.4% of all cautery occlusions preceded a cleaning. The higher proportion of occlusions preceding a cleaning for camera movement may be attributed to the surgeon's preference to clean the lens at a time in which they may be searching for a new visual angle by moving the camera, which could, in turn, be further connected to the position of the visual distortion related to pursuit of a better viewing angle. Other notable categories where a high proportion of a specific occlusion type preceded a cleaning were 'Insert/ Remove Instrument' (67.7%) and Irrigation (61.1%). Data here, given the type of active instrumentation preceding the occlusion event, may indicate that such instruments could create debris events severe enough to warrant an immediate cleaning. Conversely, this may also suggest these instruments are used at points in the surgery where the surgeon deems it more acceptable to stop the surgery and clear the lens based on their assessment of the risk threshold.
Perhaps the most intriguing finding within this study involves a comparison between average time spent cleaning the scope and time spent under suboptimal vision. Although cleaning the lens comprised 1.16% of the operative time, 41.4 ± 28.1% of all surgical observation time was spent with distorted visualization. Given that surgeons operated under suboptimal conditions for nearly 35 times as long as it would take to clear the lens, this may indicate a tendency to avoid interrupting surgical progress to clean the lens. As hinted at above, this is further supported by data from this study showing that active instrument categories such as cautery and scissors/cutting less commonly preceded a cleaning event when compared to transitional events such as camera movement and inserting/removing equipment. In other words, when using instruments for a critical part in a procedure, such as manipulation or destruction/repairing of tissue, the surgeon may be less likely to stop the procedure and clear the lens. Conversely, during less involved portions of surgery, such as when instruments needed to be swapped, a cleaning event was executed more often since the risk level to the patient was deemed lower and the need for enhanced operative focus may be deemed less critical at such a time. This is particularly important as one would surmise that surgeons prefer clear visualization in more critical moments; however, the data suggest that operating under suboptimal vision may still be preferred to the interruption and total loss of visualization that is required by a lens cleaning. This is an interesting "surgical catch-22" to consider, given the data. Anecdotal evidence via surgeon and surgical technician feedback also supports this hypothesis.
Lens occlusions were commonly discussed by surgeons and other OR personnel as a frequent inconvenience or source of frustration. One surgeon commented on his personal dislike for removing the lens to be cleaned, as it left him blind while surgical instruments were still inside the body cavity. Several surgeons and surgical residents expressed their irritation with occlusions and cleanings interrupting workflow and stated that they clean the lens as infrequently as possible to avoid that interruption. Table 3 documents raw interview response data from surgeons who shared their experience and preferences regarding lens cleaning.
Of note, there was a high degree of variability observed for operative time spent with suboptimal vision. Based upon our observations, we postulate that this is due to the personal preference of the surgeon operating, with some tending to clean immediately and often, and others operating with an occluded lens for longer. This hypothesis is supported by the interview data, which shows three approaches to lens occlusions based on the preference of the surgeon. Again, this appears to be due to frustration surrounding the prospect of interrupting surgical progress. Future research should investigate potential trends seen in different surgeons and surgical specialties.
The importance of clear visualization in RAS cannot be understated and is important for surgical success. However, this study has shown that it is impacted by a number of factors surrounding lens occlusions, including surgeon reluctance to remove and clean the scope to prevent disruption Table 3 Surgeon quotes regarding suboptimal vision Surgeon Input Surgeon 1 "I personally do not like any occlusion on my lens during laparoscopic surgery because I find it very distracting, even in situations when it is not obstructing my view to proceed with surgery." "However, in situations when there has been several attempts to clean the lens satisfactorily or if we are in a critical portion of the case that is ongoing and the occlusion does not obstruct moving forward safely, I will delay cleaning the lens until there is a better time to stop the surgery to clean the lens." Surgeon 2 "Sometimes with partial occlusion-remember one eye may be perfect while other eye bloody/ hazy-a critical step is underway." "To go blind and lose visual control of your robot arms with proximity to critical anatomy is more stressful/dangerous than continuing with suboptimal clarity." "If we had a device inside that could sweep in and clean, those 2 s would be easily tolerated and the camera view would immediately be back in desired location." Surgeon 3 "First and foremost, taking the camera out to clean is a pain and affects the flow of surgery." "It varies from surgery to surgery the dirtiness tolerance-where you're at in terms of anatomy. What are the stakes and risk of injury?" "Robotic surgery is stereoscopic so you can close one eye, making a partial occlusion more tolerable." "For example, with an automatic car wiper, you're going to have much less tolerance for a foggy windshield at night on a mountain road than you are in the middle of the day on a straight flat road; the same can be said for a difficult surgery in sensitive anatomy versus a hernia." of the surgery, based on raw interview feedback. This has potential implications for surgical efficacy and at the very least reveals a need for future research that may examine the impact of occluded visualization and lens cleaning on other aspects of surgery. Despite this study showing how profound visual disruptions are in the operating room for RAS, there are several limitations that need to be addressed. First, data collection for this study was conducted at a teaching institution where surgeons of multiple subspecialties performed RAS. All surgeons had received specialized training for RAS, but some variability in experience exists due to the participation of surgical residents during the cases. This variability is also true of the surgical technicians who handled scope cleaning. Second, a variety of procedures and surgeons were observed to gain a broad understanding of lens occlusions and cleanings across different uses of RAS. Following this study, we acknowledge that a larger scale inquiry would be able to further explore the explanations for some of the results seen. For example, variation in time spent operating with suboptimal vision between individual surgeons and variation in lens cleaning time between types of occlusions would be valuable. Third, a more formalized method of collecting surgeon interview data about visual acuity and lens cleaning could yield concrete conclusions on the impact of occlusions on their actions during a case.

Conclusion
This study provides an initial quantitative groundwork into visualization variability within the field of RAS and its ubiquitous impact across all observed surgical specialties in terms of the rate of occlusions and cleanings throughout an operation. Given the extensive amount of time surgeons spend with occluded vision, further study is warranted to more intently explore and understand the underlying causes for lens occlusion events with additional consideration to focus on consequential clinical impact. Additional avenues for research may include studies focused on variability between surgical procedures within specialties, variability between individual surgeons and surgical technicians, and degree of occlusion severity between events. These and other avenues could provide substantial evidence to support enhanced clinical decision-making surrounding these undesirable surgical events and ultimately may ideally improve surgical efficiency.
Author contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by JYD, MG and ZJR. The first draft of the manuscript was written by JYD, MG and ZJR and all authors commented on previous and subsequent versions of the manuscript. All authors read and approved the final manuscript.
Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.
Ethical approval This is an observational study. No ethical approval is required.

Data availability statement
The data that support the findings of this study are available from the corresponding author, CI, upon reasonable request.