Design and evaluation of a smart passive dynamic arm support for robotic-assisted laparoscopic surgery

(1) The device should be able to support user arms weighing between 2.8 and 5.4 kg, which corresponds to 5.3% of the weight of individuals ranging from 54 to 101 kg (average weight women −25% – average weight men + 25%) [9]. (2) The device should be strong enough to withstand a load of 11 kg on each arm support without plastic deformation, equivalent to twice the weight of a 101 kg person’s arm. (3) The device should allow the required range of motion (ROM) for using the AdLap-RS system [10], with ROM specifications as follows: For the elbow: 20 cm in the x-axis, 42 cm in y-axis, and 20 cm in the z-axis. For the wrist: 31 cm in the x-axis, 46 cm in the y-axis, and 20 cm in the z-axis. (4) The arm support should not negatively influence performance on the AdLap-VR for the parameters of time, path length, and the number of collisions. (5) The arm support should be adaptable for use in both standing and seated postures. (6) Users should be able to install and remove the arm support within 30 s. (7) When not in use, the arm support should be folded away without any hindrance, with the maximum protruding distance from the AdLap-VR case at all sides not exceeding 15 cm. (8) The positions of the points of contact of the arm support with the arms of the user should be measurable with an accuracy of 10 mm and a step sensitivity of 2 mm.

Potential design solutions were validated based on the following performance criteria. (1) Comfort, the higher the comfort level, the better the scores. (2) Cost of materials. (3) Complexity, measured by the number of moving parts, (4) Smoothness of motion during use. (5) Intuitiveness, easier setup and use result in a higher score. (6) Volume, a lower system volume during transport and use is given a higher score. (7) Durability, a maintenance friendly system that is less prone to errors receives a betters score.

In line with the BMD process for surgical devices [8], a morphological chart was created that comprises sub-solutions for partial system functions. Within the chart, three design routes were identified that can lead to a potential system. One is the most durable, one leads to the most compact design, and one to the most adjustable system. These concepts were assessed using a Harris profile and the best concept was worked out in full detail (Supplemental file 2). All structural parts were evaluated through simulations using Final Element software (Solidworks v.2022, SW corporation, Waltham, Massachusetts, USA) before being manufactured into a functional prototype. To track the opposition of the arms, new rotations sensors were developed based on the magnetic hall effect. Supplemental file 2 shows how these sensors were developed, integrated and validated in order to track the position of the arm supports. Figure 1 shows the complete system as being used during the experiments.

Biomechanical Students at the Delft University of Technology were recruited for voluntary participation in the study as novices. All participants were first shown an instructional video of the task to ensure a baseline of equal information prior to the tests. A pre-test was conducted on the AdLap-VR to familiarize participants with the system’s inputs and the digital environment. The participants were then randomized into two groups for a crossover study (Fig. 2). Both groups completed a single exercise on the AdLap-VR system for a total of eight times. The first group conducted the first four trials without arm support, followed by four trials with arm support. The second group did the first four trials with arm support, followed by four trials without arm support.

In the exercise, participants were asked to sort marbles by color into matching bowls. They were instructed to sort the green marbles with their left hand and the blue marbles with their right hand to ensure bimanual performance, and alternating left and right to ensure no differences in strategy. After completing the tests, participants filled out a questionnaire and provided feedback on their experience using the AdLap-VR with and without the arm support system. The questionnaire focused mainly on the system’s ergonomics. It included questions scored from 1 to 20 formatted according to the NASA-TLX questionnaire [11], combined with open questions regarding their experience (Supplemental file 3).

The main performance parameters from the AdLap-VR in this study included time (s) to complete the task, total instrument path length (mm), and the number of collisions. A time limit of 5 min was set for each trial for logistical reasons, and trials exceeding this time limit were marked as “did not finish” and excluded from the analysis. The subjective parameters used were mental demand, physical demand, effort, frustration, and self-perceived performance. The participants also rated the intuitiveness of the arm support, which was solely used for an indication and not for comparisons. Position data for the arm cups were derived from rotation sensors on the arm supports during trials that involved arm support.

The prototype met all the requirements. The system consisted of multiple linkages, providing four degrees of freedom at the arm pad. Rotational joints in the horizontal plane create translation along the x and y axes and rotation around the z-axis, while the four-bar mechanism is responsible for translation along the z-axis. All the linkages were constructed using aluminum square rods, and the joints consisted of ball bearings in combination with solid steel rods. All the custom-made parts at the joints were fabricated from aluminum using a milling machine.

The primary design requirement was to support users’ arms throughout the entire range of motion necessary for using the AdLap-VR. In the conceptual phase, it was determined that steel springs would be used to achieve this goal due to their durability and simplicity of use. A balanced system with adjustable support force based on a “zero-free length spring” design, as described in by Herder et al., on balancing mechanisms was integrated in the design [12, 13]. This design combines the increase in strength of the spring with a decreasing effective pulling angle on the bottom rod of the four-bar mechanism, canceling each other out when lowering the support height. This results in a mechanism with a constant support force in the working range, as opposed to a variable support force. The mechanical principle and functional parts are shown in Fig. 3. It is a requirement that the spring depicted in Fig. 3 can only be stretched and never compressed, which was achieved through the use of cables and pulleys symmetrically attached.

The support force is adjustable by rotating a knob on top of the support’s base, which, in turn, pulls the attachment point of the cable upwards via a spindle. As the height of the attachment point increases, the moment exerted on the bottom rod increases, leading to an increases in the support force. Further details on the sensor development and functioning used to tract the arm support’s movements can be found in supplemental file A3.

A total of 20 students participated in the study, resulting in 160 performed trials on the AdLap-RS. All participants completed the pre-tests. Two participants exceeded the 5-min time limit in the pre-tests, and none did so during the trials. Only one participant could not complete the second part of the trials due to external circumstances and had to be excluded from the analysis. Therefore, a replacement participant was recruited. All participants completed the questionnaire. Participants found the task, in combination with the arm support, significantly less mentally and physically demanding, requiring less effort, and resulting in less frustration. All participants expressed a high score for the intuitiveness of the device (Fig. 4).

Between the first and last trials of the task without arm support, there was a significant decrease in time, as well as between trials 1 and 3 of the task with arm support (Fig. 5).

Regarding the collision data, a significant decrease in the number of collisions was observed between trials 1 and 4. In trials 2 and 3, participants had significantly lower path lengths when using arm support compared to trials without arm support. In the final trials, participants also had significantly fewer collisions when using compared to trials without arm support. Figure 6, shows that the participants, overall, had significantly lower path length and fewer collisions in the tasks with the novel arm support as compared to having no arm support.

The movement during the arm support of all recorded trials is shown in Fig. 7. To prevent data cluttering, only the first ten participants are displayed. The variability of movement and the main directions are graphically presented as projections of oriented ellipsoids with the principal axes extending from the surface. The data shows that for the individual participant, the position of the handle in space migrate during the training task, the principal components do not change much. Between the participants position differences up to 800 mm in the horizontal plane and 700 mm in the vertical plane were observed.

Although learning effects can be observed in Fig. 5 for the three different parameters, it appears that the curve for task time shows almost no differences between the two conditions, suggesting that the presence of an arm support does not add complexity to the task execution. For path length and collisions, the parallel learning curves have a somewhat similar profile but with an offset. Significantly lower path length and collision rates were observed for the combined trials with arm support compared to the combined trials without arm support, whereas the parameter of time showed no significant differences (Fig. 6). A shorter path length with equal time indicates a reduction in participants in the average speed of the participants on the AdLap-RS [10, 14]. This reduction in speed could explain the lower number of collisions, as it suggests slower, more stable, and more precise movements. This data aligns with the majority of participants’ statements after their experience with the arm support. Moreover, in trials without arm support, participants tend to position their non-working arm towards a resting position, lowering the upper arm parallel to the upper body to reduce shoulder effort. This positional change from the working position to the resting position and vice versa creates extra path length. In contrast, during trials with arm support, participants maintain their non-working arm in a position similar to the working position, as the arm support prevents fatigue. This likely contributed to a decrease in path length for participants in trials with arm support.

Figure 7 shows that the movement areas are grouped per participant, indicating only minor adjustments in posture as participants learned and aimed to improve their performance. These adjustments were limited as participants were bound by the unchanging dimensions of their body, likely related to the height of the equilibrium point of the system arms (i.e., balancer and master arms combined). In the vertical plane (i.e., z-axis), motions were more pronounced compared to the horizontal plane, potentially due to small changes in siting position. Regarding the vertical plane, eight out of ten participants increased the height of their left arm, and nine out of ten participants increased the height of their right arm between the first and last trial with arm support (Fig. 7). This change in arm height can be explained by a development in strategy of the participants for the specific task. It was observed that grabbing the marbles from the top as sort of a crane with a grab is a more efficient way of performing the task than grabbing the marbles from the sides. Most participants intuitively learn this and moved toward this better strategy during the trials. This can explain the trend of the increasing height of the arm positions of most participants, as the more optimized strategy requires a higher position of the AdLap-VR handles, and thus a higher position of the arm supports for comfort. Another interesting observation is that the size of the ellipsoids in Fig. 7 of each participant, and the main directions of variability from the principal axes, appear to stay the same throughout the trials but are different between participants. This indicates that participants keep the same dimensional movement patterns but express them from different starting locations throughout the trials, due to slightly different postures.

When reviewing Fig. 7, the black, green, and yellow ellipsoids appear significantly lower placed than the remainder of the ellipsoids. However, upon further inspection of the data, no further conclusions can be drawn from the baseline demographics. One cause could be that both participants set the arm support strength at an insufficient level, causing the arm to find a supporting equilibrium height lower than other participants. Finally, the large operating differences between participants in general indicates the importance of dynamic arm support over a static arm rest. This makes sense as the level and location of support should not be defined by the console but by differences in body and extremity dimensions, variations in muscle strength and length of the procedure. Moreover, surgeons should be able to reposition their arms in order to prevent fatigue and overloading during the procedure.

Firstly, the system was only tested with students, not actual practicing surgeons. While this study serves as a good indicator of the design’s functionality, it is still uncertain how the arm support improves the experience of surgeons within the operating room. Further research, incorporating practicing surgeons as participants, is needed to determine the clinical utility and impact of the arm support on behavior (e.g., clutching of the handle vs instruments), surgery outcomes and ergonomics [15]. Secondly the trial times for the short marble sorting task are not representative of the extended durations of surgical operations that can last for hours. Some participants from the presented study even noted that the short trial times did not sufficiently induce fatigue, thereby limiting their ability to discern differences in comfort levels. This most likely limited differences in performance as well, as a paper by Z. Tsafrir [7] on laparoscopic performance describes that the time to complete tasks increases with increased tiredness. This effect was not found in the current study, where the parameter time showed no differences between groups. To fully investigate the impact of the designed interventions on the total comfort experience of surgeons, trial times can be increased in future studies to allow for the onset of fatigue, enhancing the applicability of the findings to real-world surgical scenarios.

To further develop the system, to automate the procedure of adjusting the support strength to the arm weight of the user with the use of force sensors, as some participants indicated that they found it difficult to find the right setting. This could be achieved by replacing the turning knob on top of the system with a small motor.