Tele-Robotic Surgery for Combat Casualty Care

Abstract

Objective:

These experiments test the feasibility of tele-robotic surgery for the shipboard care of combat casualties.

Summary Background Data:

Limited research on the effects of telecommunications latency on tele-robotics has been a consistent concern in the growth of tele-robotic surgery. These experiments are designed to provide well-controlled real-world data on robotic surgical performance with electronically added latency that simulates combat-relevant, satellite-based tele-robotic communication.

Methods:

Surgeons at 2 university hospitals separated by 1096 km bi-directionally performed remote cholecystectomy (Ch) and right colectomy (Co) dissections on cadaver subjects. Varying latency between 0 and 300 ms was electronically added. Network performance, OR time, Global Performance (GP), and Surgery Task Load Index (TLX) were determined. Statistical analysis: simple linear regression.

Results:

Network latency was 28 ms, jitter 0.03 ms, with zero packet loss. Including robotic intrinsic latency of 50 ms, total latency (TOT) varied from 78 to 378 ms. 12 Ch and 14 Co were performed by 4 surgeons. Co OR time unexpectedly decreased with TOT (P=0.005), but Ch did not (P=0.47). OR time was strongly associated with general surgeon experience (P=0.024). TLX was strongly associated with surgeon experience (P<0.001) but not TOT (P=0.47). GP was consistently between 4/5 and 5/5 across all domains and did not correlate with TOT (P=0.41). Each surgeon showed rapid cognitive and motor adaptation to increasing latency.

Conclusions:

Experienced robotic surgeons rapidly develop cognitive and motor adaptation to tele-robotic latency. Satellite-based tele-robotic surgery in combat environments may add greater latency than fiber-optic-connected civilian applications. Our data indicate a significantly greater tolerance for latency than previously reported. Development of telerobotic surgery for combat casualties is highly feasible.

Future large-scale conflicts are expected to challenge existing paradigms of combat casualty care due to anticipated delays in evacuation, constrained access to surgical teams, and increased operational demand. In response, tele-robotic surgery has been proposed as a potential force multiplier, enabling expert surgical intervention across geographic distances when local surgical expertise is unavailable or evacuation pathways are limited. In this context, a force multiplier refers to the ability to extend limited forward surgical capability by enabling remote participation of expert surgeons who are not physically deployed to the combat zone. Early demonstrations of remote robotic surgery, including the first clinical telerobotic service performed over several hundred kilometers, confirmed the feasibility of procedures such as colectomies and inguinal hernia repairs using dedicated network connections, albeit with perceptible system delays that surgeons learned to accommodate.¹

Subsequent clinical and experimental work has extended these findings across multiple specialties, including coronary intervention, multifocal laparoscopic procedures, and long-distance animal and human models, using a range of network architectures from fiber-optic virtual private networks to high-speed commercial links.^2,3 Systematic reviews of the literature highlight the breadth of procedural types and communication environments explored, while underscoring persistent concerns regarding procedural safety benchmarks and substantial methodological heterogeneity across studies.^4,5

Telecommunications latency has consistently been identified as a principal barrier to broader adoption of telesurgery.⁶ Simulator- and task-based investigations have demonstrated associations between increasing delay and degraded performance, including prolonged task completion times, increased error rates, and greater subjective workload.^7–12

Although many studies suggest that delays below ∼200 ms are generally tolerated with minimal impact on performance, measurable deterioration often emerges at higher latency thresholds, particularly during complex or dynamic tasks.¹³ Importantly, much of the existing literature relies on simulation environments or isolated task metrics that may not fully reflect operative performance under real-world network conditions.¹⁴

Technological advances in robotic platforms, data compression, and network infrastructure have substantially reduced baseline communication delays within civilian health care systems.^15–17 However, future military applications are more likely to depend on satellite-based communications, including low-Earth-orbit (LEO) systems, which introduce additional and potentially variable latency. A precise understanding of the combined effects of intrinsic robotic latency and externally imposed network delay on operative performance, surgeon workload, and procedural safety is, therefore, essential to assessing the feasibility of tele-robotic surgery in combat and austere environments.

The present study addresses this gap by evaluating tele-robotic surgical performance using a real-world, long-distance fiber-optic connection between 2 academic centers. Transoceanic, tele-robotic combat casualty care will require satellite connections to ships at sea or forward hospitals. The additional latency associated with satellite communication is a potential barrier to development. These experiments are designed to add varying amounts of electronic augmentation of latency to simulate combat-relevant satellite communication. By assessing the results of abdominal dissections in cadaveric models and measuring operative time, global performance (GP), and cognitive workload, this work aims to define practical latency tolerance thresholds and characterize surgeons’ cognitive adaptation to tele-robotic latency under conditions directly relevant to future military surgical scenarios.

MATERIALS AND METHODS

Study Design and Experimental Setup

This was a prospective experimental study evaluating tele-robotic surgical performance under controlled latency conditions using a long-distance fiber-optic network connection. A dedicated fiber-optic cable connection was established between 2 academic medical centers located 1096 km apart. Surgeons at each institution performed procedures bi-directionally on cadaveric subjects located at the opposite site using a modified Mira surgical robotic system, shown in Figure 1. The Mira is a compact, lightweight FDA-approved robotic surgical system that is amenable to shipboard and forward hospital settings.

FIGURE 1.

Modified Mira surgical robotic system supplied by Virtual Incision, including the (A) Minibot and (B) Console.

The robotic system used in these experiments was modified to reduce its intrinsic latency to ∼50 ms. The FDA-approved system without modification has an estimated latency of 70 ms (Fig. 2). Our experimental model is based on a fiber-optic cable connection between the University of Illinois Chicago (UIC) and the Uniformed Services University for Health Sciences (USUHS) in Bethesda, Maryland. Surgeons at both locations performed tele-robotic surgery on cadaver specimens located in the respective surgical laboratory located at the distant center. Electronic signals travel in a fiber-optic cable at approximately two-thirds of the speed of light (2×10⁸ m/s). Our data showed that the network latency between USUHS and UIC was consistently 28 ms (Fig. 2). The baseline latency in this system was 78 ms, including the intrinsic latency of the robot and the network latency. Variable latency between 0 and 300 ms was added electronically to simulate communication delays expected with satellite-supported tele-robotic systems. As worldwide high-bandwidth satellite communication is evolving, the latency of remote may vary depending on the location.

FIGURE 2.

Conceptual schematic of a tele-robotic surgery network spanning a 1096 km fiber-optic link between 2 academic medical centers, illustrating system architecture and latency components. A dedicated fiber-optic connection enabled surgeons to operate a modified Mira robotic system at a remote site. Total system latency (78–378 ms) was derived from the combined contributions of intrinsic robotic processing latency, baseline network latency, and electronically added delay.

Surgical Procedures

Two procedures of differing technical complexity were selected: laparoscopic cholecystectomy and right hemicolectomy dissection without anastomosis. Procedures were performed on cadaveric models acquired from Innoved Institute. All dissections were conducted using standard robotic surgical techniques. Anastomosis was intentionally excluded to isolate dissection performance and avoid confounding by suturing time. A total of 26 procedures were performed, including 12 cholecystectomies and 14 right hemicolectomy dissections.

Network Performance Monitoring

Network performance metrics, including round-trip latency, jitter, and packet loss, were continuously monitored throughout each procedure. Latency is defined as the length of time required for data to be transported from one site to another. In surgical robotics, the round-trip latency is the amount of time between initiating a movement with the handpiece and seeing the movement of the connected surgical instrument on the console display or eyepieces. Jitter is the variability of latency. The imaging data are compressed into packets for transport across the network. If packets are lost (“packet loss”), image quality may degrade. Network latencies are reported in milliseconds (ms). Packet loss is measured as the percentage of packets that do not traverse the network connection. It is measured at the network layer that sees individual packets as they depart/arrive. This information is monitored by the network equipment in near real time.

Outcome Measures

The primary outcome was operative time, defined as the total time from skin incision to completion of dissection. Secondary outcomes included technical performance and cognitive workload. Technical performance was assessed using GP scores assigned by a senior attending surgeon on a validated 5-point scale evaluating respect for tissue, time and motion, instrument handling, knowledge of instruments, flow of operation, use of assistants, and knowledge of the specific procedure.¹⁸ The senior surgeon who scored GP was aware of the amount of electronically added latency. Subjective cognitive workload was assessed immediately after each procedure using the NASA Task Load Index (NASA-TLX), a validated multidimensional instrument that measures mental demand, physical demand, temporal demand, task complexity, situational stress, and distractions.¹⁹ Surgeon experience variables included years of general surgical experience and years of robotic surgical experience.

Statistical Analysis

Continuous variables are reported as mean±SD unless otherwise specified. Associations between total system latency and operative time, GP, and NASA-TLX scores were assessed using simple linear regression. Multivariable linear regression models were constructed to assess the independent effects of latency and surgeon experience on operative time. Paired comparisons were used to evaluate learning effects across sequential cases for individual surgeons (paired t test). Mixed-effects modeling was used to account for repeated measures by each surgeon, with the surgeon treated as a random effect. Statistical significance was defined as P<0.05. Analyses were performed using RStudio [version (2026.01.0+392)] and SPSS [version (31.0.0)].

RESULTS

Network Performance and Latency

The fiber-optic network demonstrated high stability throughout all experiments, with a mean network latency of 28 to 29 ms, jitter of 0.03 ms, and negligible packet loss. Total system latency ranged from 78 to 378 ms, depending on the electronically added (0–300 ms) delay.

Operative Time

For cholecystectomy, operative time did not vary significantly with increasing total system latency (simple linear regression, P=0.47; Fig. 3). Cholecystectomy operating times ranged from 13 to 49 minutes. In contrast, operative time for right hemicolectomy dissections was significantly associated with total latency (simple linear regression, P=0.005; Fig. 3), with latency accounting for a substantial proportion of variability in operative duration. Colectomy operating times ranged from 20 to 63 minutes. In a linear mixed-effects model fit by maximum likelihood with surgeon included as a random effect, total communication latency was a significant predictor of operative time (type III ANOVA, P=0.036). The magnitude of the effect was small, and operative times remained clinically acceptable across all latency conditions. Latency was associated with a small decrease in operative time (β=−0.05 min/ms), consistent with rapid surgeon adaptation. Though this finding is counterintuitive, we observed rapid cognitive adaptation to operating in a higher latency environment.

FIGURE 3.

Relationship between total system latency and operative time during tele-robotic abdominal procedures. Operative time (in minutes) is shown as a function of total system latency (ms) for cholecystectomy and right hemicolectomy. Each point represents an individual case. Lines indicate fitted regression relationships. While operative time for cholecystectomy was not significantly associated with latency, operative time for right hemicolectomy demonstrated a significant association with increasing latency.

Surgeon Experience

General surgical experience was significantly associated with shorter operative time (simple linear regression, P=0.024), whereas robotic-specific experience was not (P=0.36). In multivariable analysis including both experience variables and latency, only general surgical experience remained independently associated with operative duration (P=0.043).

Cognitive Workload (NASA-TLX)

NASA-TLX scores were not associated with total system latency (simple linear regression, P=0.47; Fig. 4). In contrast, TLX scores were strongly associated with both general surgical experience (P<0.001) and robotic experience (P<0.001), indicating that more experienced surgeons perceived a lower workload. Analysis of individual TLX subscales demonstrated similar patterns, with no latency-dependent increases in mental demand, physical demand, temporal demand, task complexity, situational stress, or distractions.

FIGURE 4.

Surgeon cognitive workload as measured by the NASA-TLX across total system latency conditions. Composite TLX scores are shown as a function of total latency (ms). Each point represents an individual procedure. No significant association was observed between total system latency and overall cognitive workload.

Global Performance

GP scores remained consistently high across all latency conditions, with scores ranging from 4/5 to 5/5 across all evaluated domains. No association was observed between GP and total latency (simple linear regression, P=0.41).

Learning Effects and Adverse Events

Comparison of early versus late cases for individual surgeons demonstrated nonsignificant reductions in operative time for both cholecystectomy (paired t test, P=0.058) and right hemicolectomy (paired t test, P=0.071), consistent with rapid adaptation to increasing latency (Table 1). One bile duct injury occurred during cholecystectomy in a poorly preserved cadaver and was not attributable to deviation from standard dissection technique or latency effects. No latency-related adverse events were observed.

TABLE 1.

Paired Comparison of Operative Time Between Early and Late Cases for Cholecystectomy and Right Hemicolectomy

Comparison of Operative Time Between Early and Late Cases by Procedure
Procedure	Early Operative Time (Min)	Late Operative Time (Min)	Mean difference (Late-Early)	Paried T test (P)
Cholecystectomy	37.8±8.5	28.7±9.0	−10.3	0.058
Colectomy	50.0±14.7	28.8±9.0	−21.2	0.071

Trends toward reduced operative time were observed for both procedures, though differences did not reach statistical significance.

DISCUSSION

This study provides controlled, real-world evidence that tele-robotic surgery remains technically feasible and cognitively tolerable across a range of communication latencies relevant to future military and austere-care environments. Importantly, these data demonstrate a substantially greater tolerance for telecommunication delay than has been suggested by prior simulation-based or single-task studies, particularly when procedures are performed by experienced surgeons using stable network infrastructure. Caution in interpreting this data is appropriate since this was not a live animal model and did not have the risk of bleeding or the risk of nonsurvival because of surgical error. The experiment was designed to find an upper limit of tolerable latency by increasing the electronically added latency to 300 ms, a level we anticipated would be a point of failure. Our observation that this level of latency was tolerable was unexpected. While these data are encouraging, these findings should be interpreted with caution.

Telecommunication latency is a generally unnoticed but common experience in everyday life. During the evening, when Internet use is much heavier, many consumers may notice delayed initiation of downloads and uploads. Tests of Internet connectivity during heavy evening use may show prolonged “ping” or latency in telecommunications connections. Some consumers may notice that delays >100 ms are a nuisance during chats and voice-over-Internet calls. Computer game players have a very low tolerance for latency and notice diminished gaming experience with as little as 50 ms of latency. Thus, some consumers, especially computer gamers, use the Internet after midnight to optimize their experience. Though everyday experience with telecommunications latency cannot be commingled with the critical nature of tele-robotic surgery latency, it is possible that these ordinary day-to-day experiences may unconsciously influence opinions on safe tele-robotic latency maximums.

Because of concerns about managing hemorrhage, many authors have concluded that a maximum tolerable latency in tele-robotics should be much lower than was achieved in these cadaver experiments. Nankaku and colleagues observed modest increases in task completion times with increasing latency for students, residents, and experienced surgeons performing simple robotic tasks such as the movement of sticks and rings. From this data, Nankaku et al¹⁴ concluded that telerobotic latency should not exceed 100 ms. Before the Lindbergh case, Marescaux and colleagues reported a porcine experiment in which latency varying from 20 to 551 ms was electronically added. From that data, he concluded that the upper limit of tolerable latency was 330 ms; however, the details of those experiments are not provided.^5,20 Wang and colleagues have reported the positive outcome of 108 live porcine surgical procedures, including nephrectomy, oophorectomy, bladder rupture repair, and successful management of a model of bleeding vena cava repair. Those experiments were reported to have a latency that varied from 170 to 320 ms.²¹ Our experimental model was designed to address the maximum tolerable latency for 2 common shipboard surgical procedures. Next steps will require successful repetition of these cadaver experiments in a live porcine survival model. These experiments should include models of management of arterial and venous bleeding in varying latency environments. Nevertheless, from a practical standpoint and as a matter of patient safety, initial clinical trials of robotic telesurgery should require the on-site availability of a surgeon capable of responding to intraoperative complications that cannot be managed remotely.

Connection to a shipboard or a remote forward hospital will require the use of a LEO satellite connection. Initially, during the design of these experiments, we assumed that connection in the western Pacific would be by satellite only, an arrangement that would require a long latency that might be associated with multiple satellite-to-satellite links. On the basis of that assumption, our experimental model was designed to simulate the ability to perform surgery with a long latency period associated with communication through satellite only, like the long latency (∼600 ms) experiments performed on the International Space Station.²² Our more recent operational planning proposes to use a long-distance transoceanic fiber-optic cable bridged to shipboard by a LEO satellite connection. In this arrangement, a fiber-optic cable connection from Walter Reed National Military Medical Center to the western United States would connect to the western Pacific through a transoceanic fiber-optic cable to the land point closest to the point of Pacific combat operations. For example, a connection to a ship in the South China Sea would be by fiber-optic cable to Japan (10,870 km). Round-trip connection time at 2/3 the speed of light in fiber-optic cable (4.9 µs/km) would be 106 ms. LEO satellite connection from Japan to the shipboard nearby is estimated to add 20 to 80 ms (Pete Bohley, Sovato Health, personal communication). The cumulative fiber-optic cable plus LEO satellite connection to the western Pacific is estimated to have a latency of ∼126 to 186 ms. Another technical issue with a LEO satellite connection may be the satellite-to-satellite handoff that occurs as the satellites revolve around the Earth. Each LEO satellite has a cone of operational connectivity of about 150 km². Continuous coverage is maintained through an array of satellites that hand off to the next satellite reaching the cone of coverage. This may cause a momentary glitch in connectivity when the satellite-to-satellite handoff occurs. Our future porcine experiments will test these real-world satellite communication dynamics. In this scenario, surgeons may be required to pause for the brief satellite handoff and be prepared to experience latency that varies intraoperatively. The experiments in this report were not designed to assess intraoperative latency variation; future porcine live animal experiments will test this potential risk. This is a field that is undergoing rapid technical evolution that, at times, is at a faster pace than the surgical experiments can be designed and executed.

Our data showed that procedure complexity appeared to influence the relationship between latency and operative duration. While operative time for cholecystectomy was not significantly associated with latency, right hemicolectomy dissections demonstrated a statistically significant inverse association between latency and operative duration. The direction and small magnitude of this association suggest adaptive changes in surgeons’ technique. The surgeons were unfamiliar with this robotic system; however, as they became increasingly facile using the robot, operative times became shorter for the colonic dissection. We also observed that the surgeons recognized there was no risk of bleeding in the mesentery; the dissections began to proceed faster in some later experiments. We do not believe operative times would decrease with increasing latency; our data were an epiphenomenon of surgeons’ increasing familiarity with the robotic system and the task. In a live animal or clinical scenario, the opposite would likely be true. This divergence may reflect differences in task structure and workflow demands between procedures, as more complex operations require sustained bimanual coordination and continuous tissue interaction, potentially prompting earlier or more deliberate compensation strategies under conditions with delayed visual feedback. Notably, even under the highest latency conditions tested, effective technical execution remained excellent, and no latency-associated degradation in GP was observed. The senior surgeon observer noted that the surgeon’s robotic arm motion was slower and more deliberate, with fewer preparatory, nonproductive motions. There was a striking absence of practice motions commonly seen among less experienced surgeons. Robotic arm motions were fewer and more precise. The parametric data indicate significant cognitive adaptation to latency, particularly among senior surgeons. Nonparametric observations match this finding. During higher latency tasks, surgeon motion was more conservative and economical.

Surgeon-related factors exerted a greater influence on performance than system latency itself. General surgical experience, rather than robotic-specific experience, was the dominant predictor of operative efficiency. This finding suggests that foundational surgical judgment, anticipation, tissue-handling, and instrument-handling skills may mitigate the operational effects of delayed feedback more effectively than robotic platform familiarity alone. Cognitive workload was similarly associated with surgeon experience but not latency, reinforcing the concept that surgeons rapidly adapt internal control strategies when operating under altered feedback conditions. Notably, the senior surgeon observed sensitivity to effective communication between the surgeon and the first assistant in high-latency conditions. This was attributed to the human factor; communication breakdowns due to language barriers, ambient noise, and poor room acoustics occasionally disrupted workflow. In 2 cases, the first assistant could not hear the surgeon clearly, or vice versa, which caused the operation to come to a temporary stop. These observations highlight that effective communication and environmental awareness are essential components of surgical performance, particularly when technical latency challenges are present. Building on these observations, future work will specifically investigate how human factors influence performance under high-latency conditions. Our future experiments will include measures of the effect of remote surgery and latency on team communication and performance. Designing objective measures of team communication and performance will be a challenge for future studies.

The absence of a relationship between latency and GP scores is particularly relevant for combat casualty care. In environments where evacuation delays, resource limitations, and casualty volume may necessitate remote surgical intervention, preservation of technical quality is essential. The present findings suggest that moderate latency, including that anticipated with LEO satellite communication, does not inherently compromise surgical safety when procedures are performed by appropriately trained and experienced surgeons. Together with the stable NASA-TLX scores observed across latency conditions, these results indicate that surgeons compensate for communication delay primarily through temporal adjustments rather than increased cognitive burden or compromised technique.

CONCLUSIONS

From a military perspective, these findings have important implications. In environments where evacuation delays and casualty volume may necessitate remote intervention, preservation of technical quality is more critical than marginal increases in operative time. The demonstrated feasibility of tele-robotic surgery at latencies consistent with LEO satellite communication supports continued development and operational evaluation of these systems for combat casualty care.

This study has some notable limitations. An important limitation of these experiments is the bloodless environment of cadaver surgery. Cadaveric models do not replicate the physiological consequences of bleeding, hemodynamic instability, or time-critical hemostasis, and the absence of anastomosis limits generalizability to reconstructive phases of surgery. In addition, the sample size limits the detection of small performance effects. Next steps will include confirmation of our findings in a preclinical porcine model that assesses the ability to perform surgical interventions, such as intestinal resection, common in the postresuscitation phase of trauma care. Subsequent experiments will directly test the capacity to perform survival surgery through a satellite-supported network. We have taken an iterative approach to build an experimental case for tele-robotic trauma care. We see 3 incremental steps to make this a clinical reality: first, porcine experiments supported by a fiber-optic cable with added latency; second, porcine experiments over a satellite-connected transoceanic network; and then a phase I/II clinical trial. It should be noted that our experimental work does not consider civilian tele-robotic surgery. The case for civilian applications over a fiber-optic network is supported by recent compelling studies.^23,24

Our data demonstrate significant tolerance to moderate communication latency, with preserved technical performance and manageable cognitive workload. With stable networks, appropriate training, and surgeon experience, tele-robotic surgical care appears both technically and operationally feasible for future civilian, military, and austere surgical environments. These data support continued development and operational testing of tele-robotic systems as a potential force multiplier for combat casualty care. We conclude that the development of a satellite-based network for remote combat casualty care is feasible.

DISCUSSANT

Dr. Matthew D. Neal (Pittsburgh, PA)

This study addresses a contemporary and very important problem: how to deliver surgical care when distance, conflict, or infrastructure separates the surgeon from the patient.

In this case, over 1,000 kilometers separated the surgeon and the cadaver undergoing robotic surgery.

The authors provide convincing evidence that experienced surgeons can adapt to levels of latency that were previously thought to be prohibitive while maintaining performance and without an apparent increase in cognitive workload.

This is a meaningful and provocative contribution, particularly in the context of combat casualty care and other austere environments.

Now, I will admit, as someone who impatiently and immediately hits refresh the moment the Wi-Fi slows even slightly, that I have a great deal of respect for the surgeons who operated patiently and effectively with this degree of delay.

But perhaps the findings echo a principle taught to us by the military and that we all as surgeons recognize: “slow is smooth and smooth is fast.”

The observation that performance may actually improve with increased latency is both counterintuitive and thought-provoking, and likely reflects this more deliberate and efficient operative approach.

I have 2 questions that highlight what I believe to be 2 of the major limitations of this work.

The first is, why was a fixed latency model chosen, where the latency was maintained the same during the case, rather than varying the latency within an individual case, which may better simulate real-world communication disruptions and may have been more disorienting for the surgeon?

The second is to follow up on your future directions. How do you plan to study the impact of latency on the management of intraoperative complications, particularly hemorrhage, in the more physiological models that you suggest?

Dr. Stephen T. Bartlett (Chicago, IL)

When we initially picked the degrees of latency, we were quite sure that we were going to hit a fail point at 300 ms latency.

In fact, the surgeons found a way around that, and that was what we called the grand pause. Instead of 1, 2, or 3 preparatory robotic arm movements and then the final surgical action that you see with young robotic surgeons, they only made a single very deliberate movement, and then they stopped and waited for the action to catch up. We refer to that as a “latency pause.” Senior surgeons who are training robotic surgery residents and fellows often observe practice motions before the definitive motion. In the high-latency cases, the senior surgeon eliminated the practice motions and made one definitive motion followed by a pause.

Changing the latency period, I think it would be logical in satellite-based communication; almost certainly, when one satellite hands off to the next, there’s going to be a glitch. People were talking about that earlier in this meeting, probably a 3- to 5-minute glitch. So I think you’re very wise to recommend incorporating varying latency within the same case in our future porcine survival experiments.

The second question regarding the simulation of hemorrhage to detect the impact of latency.

One of our experiments is going to be a nephrectomy, a nephrectomy for cancer, and then to determine whether there is more difficulty accomplishing that task at certain latency levels. We agree with you that in a model where hemorrhage is a risk, we may encounter issues with control of bleeding during cases with higher bleeding risk, such as nephrectomy, but we won’t know for sure till we get there experimentally. We believe nephrectomy for trauma will be one of the models we pursue to better simulate authentic trauma conditions.

I might point out that the most common shipboard surgery on the Gerald Ford is cholecystectomy, followed by hernia. Those cases are obviously not all traumatic injuries. But nontrauma general surgical problems can remove a warfighter from duty just as much as traumatic injury. Whether you can return those soldiers to duty within 3 or 4 days, rather than taking them off the boat, has a major force multiplier effect.

Dr. Dmitry Oleynikov (Long Branch, NJ)

I was a co-investigator on this, and my disclosure is that I’m a co-founder of Virtual Incision, which seeks to commercialize miniature robots for clinical applications.

I want to speak to 2 issues that Dr. Bartlett mentioned. One is latency. In a recent experiment, we were able to launch a miniature robot onto the space station on the ISS and perform some surgery, both using autonomous robotic technology to cut some rubber bands.

Of course, we then had 6 surgeons, including myself, perform some cutting of the rubber bands themselves. The latency got as high as 850 ms. It’s a little bit like having a sort of slow motion.

The latency is felt more in one’s movements than they are in the video that you receive. So the video tends to come in a little bit crisper than the actual motion of the robot. So you have to sort of marry the 2. So it takes a little bit of effort to do so.

What we’ve determined is that, believe it or not, the expert surgeon, as you increase the latency, will become more efficient in their movement because they won’t be sort of fishing for the spot to go to. They’ll just slowly approach it.

It’s a paradigm that’s very interesting to think about, that the experienced surgeon will sort of adjust up to 850 ms, which is a long time.

My comment is that robotic surgery is here to stay, and it’s going to continue to increase.

As in some of the robotic systems that we have today, there will be the ability to do telepresence. There is some experimentation in telesurgery already in foreign countries, but not in the United States at this time.

I think this is very important research that’s going to lay the foundation for the safety of the patient on the other end of the surgeon. It certainly is very important for our military and for NASA.

Dr. Stephen T. Bartlett (Chicago, IL)

Dmitry, you were very instrumental in the success of our research project. The work you did on the International Space Station, or ISS, as you referred to it, demonstrated that certain tasks can be performed effectively even at a very high latency.

I mentioned very briefly in my introductory slide that task-based experiments, such as robotic movement of pegs and rings, can be completed successfully even when performed by students and residents under high-latency conditions. Those experiments showed good accuracy of the robotic motion, but the time to task completion increased. The work you did on the ISS showed similar results—the task could be done despite very high latency. Our future work in a porcine model will raise the bar on the impact of latency.

Dr. Eric Elster (Bethesda, MD)

Having been that surgeon a long time ago on that ship all by myself, alone and unafraid, one thought I had as I listened is, should we start to think about expanding this beyond the abdomen?

One of the issues that we have is access to neurosurgery or other specialists. So should we, as we expand the effort here, bring the robot, the abdomen into the thoracic cavity, head, and neck, or beyond?

Dr. Stephen T. Bartlett (Chicago, IL)

This is a very good idea. There’s a very good paper in the 2025 Journal of Thoracic Surgery on the successful removal of a lesion in a pig lung that was done tele-robotically, and the pig had an excellent recovery.

Application of robotic surgery to the brain, the thoracic cavity, and the limbs are, I think, possibilities that we can adapt to tele-robotic surgery for the management of a traumatized victim in a forward hospital or a shipboard.

Thank you for bringing into the discussion your real-life combat surgical experience that you gained from multiple deployments.

Dr. Nancy Perrier (Houston, TX)

Do you have any data on the eye-hand coordination of the surgeons performing the operation, such as the processing, convergence, and divergence?

Can your system measure the delay of the eye-hand process and whether there are measurable individual factors per surgeon or skill that have an effect on outcomes?

Dr. Stephen T. Bartlett (Chicago, IL)

Dr. Perrier, you are correct to focus on the cognitive activity of the surgeon. After observing many hours of tele-robotic surgery, it is clear that total surgical time includes very brief pauses for the surgeon to make cognitive decisions about the next surgical action. We are all familiar with watching residents and young surgeons make nonproductive practice motions before taking the definitive action.

After watching hours of robotic surgery and reviewing the video content, I conclude that the practice motions are a way for a young surgeon to gain time to think and process before making the definitive robotic motion. More experienced surgeons are processing information faster, in my view. Senior surgeons do not need the dilatory practice motions; they make one definitive motion followed by the “latency pause.” The ability to handle latency is a matter of executive functioning. This observation was among the valuable nonparametric data we gained from this study.

Dr. Nancy Perrier (Houston, TX)

That point is the motion. But before that, it’s eye-hand coordination that happens with us as individuals and that process of what is being seen with that delay. I wonder how that affects the cognitive function of the eyes actually converging, because they have to see it, right, before the hand motion happens. It’s an interesting concept.