Annals of Biomedical Engineering, Vol. 46, No. 10, October 2018 ( 2018) pp. 1621–1636 https://doi.org/10.1007/s10439-018-2049-z Medical Robotics Robotic Surgery Improves Technical Performance and Enhances Prefrontal Activation During High Temporal Demand 1,2,4 2 2 2 HARSIMRAT SINGH, HEMEL N. MODI, SAMRIDDHA RANJAN, JAMES W. R. DILLEY, 3 1 1,2 1,2 DIMITRIOS AIRANTZIS, GUANG-ZHONG YANG, ARA DARZI, and DANIEL R. LEFF 1 2 Hamlyn Centre for Robotic Surgery, Imperial College London, London, UK; Department of Surgery and Cancer, Imperial College London, London, UK; Institute for Liver and Digestive Health (ILDH), University College London, London, UK; and Department of Surgery and Cancer, St Mary’s Hospital, 2nd Floor, Paterson Wing, Praed Street, London W2 1NY, UK (Received 27 January 2018; accepted 11 May 2018; published online 4 June 2018) Associate Editor Cameron N. Riviere oversaw the review of this article. Abstract—Robotic surgery may improve technical perfor- INTRODUCTION mance and reduce mental demands compared to laparoscopic surgery. However, no studies have directly compared the impact The propagation of robotic techniques has revolu- of robotic and laparoscopic techniques on surgeons’ brain tionised minimal access surgery by addressing some of function. This study aimed to assess the effect of the operative 20 the limitations of the laparoscopic approach. Ro- platform (robotic surgery or conventional laparoscopy) on botic technologies do not replace the surgeon nor prefrontal cortical activation during a suturing task performed perform tasks independently, but rather provide under temporal demand. Eight surgeons (mean age ± SD = 34.5 ± 2.9 years, male:female ratio = 7:1) performed an complementary capabilities that enhance dexterity and intracorporeal suturing task in a self-paced manner and under improve ergonomic efﬁciency. Since they are con- a 2 min time restriction using conventional laparoscopic and trolled by the surgeon, they are often described as robotic techniques. Prefrontal activation was assessed using ‘master–slave systems’ and composed of two compo- near-infrared spectroscopy, subjective workload was captured nents: (1) the master console which is the user interface using SURG-TLX questionnaires, and a continuous heart rate monitor measured systemic stress responses. Task progression that provides the surgeon with a 3-dimensional view of scores (au), error scores(au), leak volumes(mL)and knot tensile the operating ﬁeld, manipulators which allow the sur- strengths (N) provided objective assessment of technical per- geon to remotely control instruments, and a control formance. Under time pressure, robotic suturing led to panel allowing adjustment of camera focus and posi- improved technical performance (median task progression tion; and (2) the slave unit positioned at the patient’s score: laparoscopic suturing = 4.5 vs. robotic suturing = 5.0; z = 2 2.107, p = 0.035; median error score: laparoscopic side on which the camera and instruments are docked suturing = 3.0 mm vs. robotic suturing = 2.1 mm; and manipulated on robotic arms (Fig. 1a). z = 2 2.488, p = 0.013). Compared to laparoscopic suturing, Robotic-assisted surgery has been increasingly em- greater prefrontal activation was identiﬁed in seven channels ployed in a number of complex procedures, particu- located primarily in lateral prefrontal regions. These results larly those in which the operative ﬁeld is conﬁned such suggest that robotic surgery improves performance during high workload conditions and is associated with enhanced activation as colorectal, gynaecological, and urological surg- in regions of attention, concentration and task engagement. eries. The growth of robotic platforms in surgery is supported by a wealth of literature demonstrating its Keywords—Neuroimaging, Brain function, Stress, Cognitive advantages over conventional laparoscopy. Robotic workload, Surgical skills, Laparoscopy, Suturing. instruments have six degrees of freedom, greater than conventional laparoscopic instruments, which im- proves instrument manipulation. Software ﬁlters min- Address correspondence to Harsimrat Singh, Department of imize hand tremors normally ampliﬁed in conventional Surgery and Cancer, St Mary’s Hospital, 2nd Floor, Paterson Wing, Praed Street, London W2 1NY, UK. Electronic mail: laparoscopy. Finally, motion scaling allows large hand firstname.lastname@example.org movements at the master console to be translated to Harsimrat Singh and Hemel N. Modi have contributed equally to micromovements of the instruments within the patient. this work. 0090-6964/18/1000-1621/0 2018 The Author(s) 1622 SINGH et al. A 3-dimensional ﬁeld of view improves depth percep- work demonstrated that intraoperative temporal stress tion which, coupled with an absence of a fulcrum ef- prompts attenuated prefrontal activations and techni- fect, makes instrument movement more intuitive and cal performance deterioration during a laparoscopic improves hand–eye coordination. Robotic platforms suturing task. However, there are no studies report- are ergonomically efﬁcient and reduce physical burden ing prefrontal responses during robotic surgery hence on the surgeon as the operator is seated comfortably at the impact of robotics on prefrontal function under a remote master console in contrast to the awkward mental demands remains unknown. positions frequently required during laparoscopic Several studies from outside medicine have shown surgery. In contrast to laparoscopy, camera motion that more mentally demanding conditions are associ- in robotic surgery is steady and controlled by the pri- ated with less prefrontal activation and a decline in 1,13,16,26,34 mary surgeon via the master console system. Further- task performance. In particular, several more, simulation studies demonstrate the robotic fNIRS studies demonstrate tasks performed under approach improves smoothness of movement, re- stress or heightened cognitive load impact adversely on duces instrument pathlength and errors of commis- prefrontal responses, including n-back working mem- 31,37 26 sion, and leads to learning curve attenuation ory tasks, mental arithmetic tasks performed under 10,37 1 compared to laparoscopy. time pressure and with negative feedback, multitask- Several studies have investigated surgeons’ mental ing during naval air warfare management, and sim- demands and have demonstrated robotic surgery to be ulated piloting tasks under variable demand. less cognitively demanding and/or stressful than la- Together, these studies suggest that conditions in paroscopy based on both subjective question- which mental demands are greatest lead to reduced 31,37,40 31,40 naires as well as cardiovascular parameters. attention, manifest as diminished prefrontal engage- For example, Moore et al., utilised the Surgical Task ment, and a deterioration in task performance. Con- Load Index, the Rating Scale for Mental Effort, and versely, less cognitively demanding conditions are assessment of heart rate variability to demonstrate associated with greater prefrontal activation and im- robotic-assisted surgery to be less stressful than the proved task engagement. Given that as discussed, ro- laparoscopy during ball pick-and-drop and rope botic platforms are found to ofﬂoad mental demands 31 14 14,31,37,40 threading tasks. Similarly, Hubert et al., observed versus laparoscopy and highly stressful or that physical workload (electromyography), stress demanding operative conditions are associated with 1,13,16,26,34 (heart rate) and perceived mental effort (NASA Task attenuated prefrontal responses, we hypoth- Load Index) were greater during laparoscopy com- esise that, compared to conventional laparoscopy, ro- pared to robotic-assisted tasks. Since intraoperative botic surgery will result in greater prefrontal activation stress and high cognitive workload can be detrimental indicative of greater task engagement and attentional to surgical performance, the suggestion that robotic control, and improved technical performance when platforms ofﬂoad mental demands compared to la- suturing under time pressure. The prediction being that paroscopy is appealing. However, to date, no studies improved instrument ergonomics and alleviation of have directly objectively assessed the impact of robotic workload is manifest as improved technical perfor- technology on surgeons’ brain function as means of mance and enhanced activation within executive con- conﬁrming or refuting this hypothesis. trol centres. Therefore, the aim of the current study is Neuroimaging has been used to assess brain func- to compare the impact of robotic surgery and con- tion during mentally demanding tasks, such as working ventional laparoscopy on prefrontal activation in sur- 5,6 memory tasks and simulated command and control geons performing an intracorporeal suturing task 6,16 tasks. The majority of this literature places under temporal demand. emphasis on the prefrontal cortex (PFC)—an area of the brain important for executive control, attention and task engagement —and suggests an inverted-U MATERIALS AND METHODS shaped relationship between PFC activation and mental workload, whereby activation initially increases Subjects with workload, but then diminishes at excessive levels 13,16 of mental demand. The use of brain imaging Following local ethical approval (LREC: 05/Q0403/ techniques to investigate surgeons’ cognition is rapidly 142), 102 surgeons from across London were invited to expanding and has deepened our understanding of enrol in this study. Subjects were only eligible to par- technical skill acquisition, decision-making pro- ticipate if they were higher surgical trainees or con- 24 23 cesses and the effects of fatigue. Although there is sultants. Eight surgeons (1 consultant, 7 trainees) limited evidence on the impact of intraoperative men- agreed to take part (mean age ± SD = 34.5 ± 2.9 tal demands on operator brain function, our previous years, male:female ratio = 7:1) and gave written in- Robotic Surgery and Brain Function 1623 formed consent prior to the study commencing. Sub- sure both HbO and HHb changes across 24 prefrontal jects had signiﬁcantly greater previous experience with cortical locations (‘channels’), the positions of which laparoscopic suturing compared with robotic suturing were deﬁned according to the international 10–20 [median number of times previously performed (IQR): system of probe placement. laparoscopic suturing = 12.5 (10.0–22.5), robotic suturing = 1.5 (0.0–3.8); p = 0.012]. Participants Stress and Technical Performance were screened for handedness [median handedness (range) = 0.80 (2 0.20 to 1.00)] and neuropsychiatric The Surgical Task Load Index (SURG-TLX) illness (n = 0), and were asked to refrain from con- questionnaire was used to evaluate subjective work- suming alcohol or caffeine for the preceding 24 h. load. Heart rate (HR) was continuously recorded using a wireless monitor (Bioharness, Zephyr Technology, USA) to provide a physiological measure of stress. Suturing Task and Experimental Paradigm Technical skill was objectively assessed using four performance parameters described in our previous The experiment was conducted in a block design in work: which each participant performed the surgical task ﬁve times in each condition (order randomized) on both Task Progression Score (TPS; Arbitrary Units, au) surgical platforms with an inter-knot rest period of 30 s. The surgical task involved intracorporeal sutur- Each task episode was assigned a score based on ing with a 2-0 Vicryl suture (Ethicon, Somerville, NJ) task progression, with one point awarded for each of inserted as close as possible to pre-marked entry and the following steps: (1) mounting the needle onto the exit points on either side of a defect in a Penrose drain. needle holder, (2) needle insertion into the drain, (3) Tying a knot required formulation of one double exiting the needle from the drain, (4) double throw, (5) throw followed by two single throws of the suture. 1st single throw, and (6) 2nd single throw of a Each subject executed the experimental paradigm ﬁrst laparoscopic reef knot. The TPS comprised the total using a conventional laparoscopic approach on a box number of points obtained during the task (maximum trainer (iSim2, iSurgicals, UK) and, after a washout score = 6). period of 6 months, repeated the experiment via robotically-assisted technique using a da Vinci Si Error Score (ES; mm) system (Intuitive Surgical Inc, Sunnyvale, CA). All Adapted from the FLS scoring system for intra- participants performed the task under two experi- corporeal suturing, the ES was calculated as follows: mental conditions on each platform, as follows: (1) Error Score = [distance (mm) between needle inser- ‘self-paced’ in which subjects took as much time as they tion point and pre-marked target position + distance needed to tie a knot, and (2) ‘time pressure’ where a (mm) between needle exit point and pre-marked target time limit of 2 min per knot was imposed (Fig. 1). position]. Accurate needle placement in vivo is neces- During the self-paced condition, the task episode was sary as else there is risk of damage to surrounding terminated if a subject required longer than 5 min to structures. complete the knot. During the inter-trial rest periods, subjects were instructed to assume a comfort- Leak Volume (LV; mL) able seated position looking directly ahead at a blank screen and avoid thoughts relating to the task. Saline was infused through each drain at a rate of 150 drops/min controlled via a digital pump. The volume of saline leaking from the closed defect over a 1 min period was recorded to assess the quality of Neuroimaging Data defect closure. Lower leak volumes would reﬂect Functional near infrared spectroscopy (fNIRS) is a superior defect closure, analogous to improved ligation non-invasive functional neuroimaging technique which security of a bleeding vessel. measures cortical absorption of near infrared (NIR) light to estimate the local concentration changes of Knot Tensile Strength (KTS; Newtons, N) oxygenated haemoglobin (HbO ) and deoxygenated haemoglobin (HHb). The typical haemodynamic brain A bench-top tensiometer (5565 single-axis ten- activation response comprises a task-evoked increase siometer, Instron, UK) was used to quantify the tensile in HbO and a lower amplitude decrease in HHb. An strength of each tied knot. Greater knot tension fa- ETG-4000 Optical Topography System (Hitachi vours knot security, an important aspect of sound Medical Co, Japan) was used to simultaneously mea- surgical technique. 1624 SINGH et al. FIGURE 1. (a) The da Vinci Si system (Intuitive Surgical Inc, Sunnyvale, CA) consists of a master console system with which the surgeon controls a slave unit comprising robotic arms that move around ﬁxed pivot points and carry out the surgeon’s commands; (b) A bench-top box trainer (iSim2, iSurgicals, UK) used for the laparoscopic suturing task; (c) Key steps of the suturing task in which a reef knot is created: (i) mounting the needle onto the needle holder, (ii) inserting the needle into the Penrose drain as close to pre-marked target points as possible, (iii) exiting the needle out of the drain as close to pre-marked target points as possible, (iv) double throw of suture thread, (v) ﬁrst single throw, and (vi) second single throw; (d) Prefrontal activation during the task was assessed using functional near-infrared spectroscopy (fNIRS), a non-invasive neuroimaging technique, which measures differences between emitted and detected near infrared light to estimate the local concentration changes of oxygenated haemoglobin (HbO ) and deoxygenated haemoglobin (HHb), as a surrogate of brain activation. The 3D head reconstruction demonstrates the 3 3 3 arrays of optodes located over the left and right PFC, along with the positions of 24 corresponding channels (‘Ch’) that measure haemodynamic responses in an area of cortex located between an emitter (red) and detector (blue). The typical cortical haemodynamic response in channels exhibiting activation comprises an increase in HbO , a smaller decrease in HHb, and an increase in total haemoglobin (HbT = HbO + HHb). 2 Robotic Surgery and Brain Function 1625 FIGURE 1. continued Data Processing and Statistical Analysis Subjective Workload and Technical Skills Data Statistical analysis was performed using SPSS ver- Between-condition and within-condition compar- sion 23.0 (IBM Corp., Armonk, NY). A threshold isons were performed using the paired samples t test p < 0.05 was deemed statistically signiﬁcant. for parametric data (i.e., SURG-TLX and leak vol- 1626 SINGH et al. ume) and the Wilcoxon Signed Ranks test for non- correlations between heart rate and changes in oxy- parametric data (i.e., heart rate, task progression score, genated haemoglobin were performed on a subject-le- error score and knot tensile strength). vel for each platform and each condition. Neuroimaging Data RESULTS Data were processed using HOMER2, an open source MATLAB-based toolbox. High-frequency Self-Paced vs. Time Pressure (Laparoscopic Suturing) noise and electrocardiographic effects were minimized Subjective Workload and Heart Rate using a low-pass ﬁlter (0.5 Hz). Channel rejection was based on amplitude thresholding and a signal-to-noise Subjective workload was signiﬁcantly greater in the ratio of > 2. Furthermore, motion artefacts were de- TP compared to SP condition (mean SURG-TLX ± tected and corrected using spline interpolation of SD: SP = 147.6 ± 52.2, TP = 202.1 ± 46.2; optical density data. The modiﬁed Beer-Lambert law t(7) = 2 4.805, p = 0.002). However, as illustrated in was used to convert changes in light intensity into Fig. 2a and summarised in Table 1, there was no sig- HbO and HHb concentration changes. Data were niﬁcant difference in DHR between conditions then processed for statistical analysis using a bespoke (p = 0.305). analytical framework (ICNNA) and self-paced blocks were resampled to 120 s, ensuring uniformity Technical Performance with time pressure blocks, before collating a database Performance was signiﬁcantly inferior in the TP of per subject HbO and HHb values for each condi- compared to the SP condition [task progression score tion, operative platform and block. A time window of (median (IQR): SP = 6.0 (1.0) vs. TP = 4.5 (2.0); 120 s with a break delay of 10 s from task onset was z= 2 4.710, p < 0.001), error score (median (IQR): selected for this purpose. SP = 2.0 (2.0) mm vs.TP = 3.0 (3.6) mm; z = 2 3.084, p = 0.002), and knot tensile strength Identiﬁcation of Channel Activation (median (IQR): SP = 25.1 (74.4) N vs. TP = 0.0 For each operative platform (laparoscopic and ro- (15.4) N; z = 2 2.843, p = 0.004)]. As illustrated in botic) and each condition (SP and TP), channel acti- Fig. 2a and summarised in Table 1, there was no sig- vation was determined by comparing the mean niﬁcant between-condition difference in leak volume baseline rest Hb data sampled over 10 s before task (p = 0.188). onset (HbO and HHb ) with mean task Hb data 2Rest Rest sampled over 110 s starting 10 s after task onset Cortical Haemodynamic Response (HbO and HHb ) using the Wilcoxon Signed 2Task Task The magnitude of channel activation varied Ranks test. Channels displaying a statistically signiﬁ- depending on task demand, with a greater proportion cant (p < 0.05) increase in HbO and decrease in HHb of channel activation observed in SP versus TP. In the were considered ‘‘activated’’, and those showing a SP condition, twenty out of twenty-four channels signiﬁcant decrease in HbO and increase in HHb were exhibited a task-induced increase in HbO concentra- considered ‘‘deactivated’’. tion, two of which reached signiﬁcance (channels 20 and 21). Similarly, twenty channels exhibited a de- Comparison of Activation Responses crease in HHb concentration, of which eleven reached For each channel, new variables DHbO (HbO 2 2Task statistical signiﬁcance. Two channels (channels 20 and 2 HbO ) and DHHb (HHb 2 HHb ) were 2Rest Task Rest 21) demonstrated both a signiﬁcant increase in HbO computed. Using the Wilcoxon Signed Ranks test, as well as a signiﬁcant decrease in HHb. Task-induced DHbO and DHHb in each channel were compared increase in HbO was observed in twenty channels in between conditions (i.e., SP vs. TP) for each platform; the TP condition, one of which reached signiﬁcance and between platforms (i.e., laparoscopic vs. robotic) (channel 6). Fifteen channels exhibited HHb decreases, for each experimental condition. six of which were statistically signiﬁcant. No channels demonstrated a simultaneous increase in HbO and Correlation Analysis decrease in HHb. Between-condition comparisons of DHbO revealed attenuated responses in TP compared In order to assess the extent of systemic contribu- to SP in thirteen channels, located predominately in tion to the cortical haemodynamic signal, channel-wise Robotic Surgery and Brain Function 1627 FIGURE 2. (a) Subjective workload (SURG-TLX) scores, heart rate, and technical performance in self-paced (yellow bars) and time pressure (green bars) conditions during laparoscopic suturing. Error bars represent the 95% conﬁdence interval; (b) Comparison of the HbO response (DHbO ) in time pressure and self-paced conditions during laparoscopic suturing. Channels exhibiting a 2 2 smaller HbO response in the time pressure compared to the self-paced condition are blue, whereas those exhibiting a smaller response in the self-paced condition are red. Channels in which there was a signiﬁcant difference (p < 0.05, Wilcoxon Signed Ranks test) in the HbO response between conditions are outlined (black circle). Reference points of the 10–20 system of optode placement are shown in yellow (right PFC: Fp2 = source 6, F8 = source 7, F2 = source 9, FC4 = source 10; left PFC: Fp1 = source 2, F7 = source 1, F1 = source 5, FC3 = source 4). Au arbitrary units, DHR change in heart rate from rest to task, bpm beats per minute, mm millimetres, mL millilitres, N Newtons. 1628 SINGH et al. the bilateral VLPFC and DLPFC (Fig. 2b). Analysis of DHHb responses demonstrated diminished responses in TP compared to SP in seventeen channels. Self-Paced vs. Time Pressure (Robotic Suturing) Subjective Workload and Heart Rate Subjective workload was signiﬁcantly higher in the TP compared to SP condition (mean SURG-TLX ± SD: SP = 115.9 ± 72.1 vs. TP = 183.9 ± 75.9; t(7) = 2 2.881, p = 0.024). As Fig. 3a highlights there was no signiﬁcant difference in DHR between conditions (p = 0.364) (Table 1). Technical Performance Under TP, there was a signiﬁcant deterioration in technical skills [task progression score (median (IQR): SP = 6.0 (0.0) vs. TP = 5.0 (2.0); z = -3.767, p < 0.001), leak volume (mean ± SD: SP = 15.5 ± 1.3 mL vs. TP = 18.3 ± 1.2 mL; t(7) = 2 8.712, p < 0.001), and knot tensile strength (median (IQR): SP = 21.2 (52.9) N vs. TP = 1.4 (13.1) N; z = 2 4.982, p < 0.001)]. As illustrated in Fig. 3a, no signiﬁcant between-condition difference in error score was observed (p = 0.451) (Table 1). Cortical Haemodynamic Response In the SP condition, ten channels exhibited task- related HbO increases, primarily in the bilateral DLPFC, and eight channels demonstrated a non-sig- niﬁcant decrease in HHb. No channels exhibited both an increase in HbO and decrease in HHb. In the TP condition, ﬁfteen channels exhibited task-induced HbO increases across the bilateral VLPFC and right DLPFC, and seven channels demonstrated a decrease in HHb, one of which reached statistical signiﬁcance. As in the SP condition, no channels exhibited an in- crease in HbO along with a decrease in HHb. Fur- thermore, contrary to the results observed during laparoscopy, DHbO was greater in TP compared to the SP condition in ﬁfteen channels located in the bilateral VLPFC (Fig. 3b). Similarly, the magnitude of the DHHb response was greater in TP versus SP in eighteen channels. Laparoscopy vs. Robotic Surgery (Self-Paced) Subjective Workload and Heart Rate SURG-TLX scores were lower during robotic suturing than laparoscopic suturing, but the diﬀerence was not statistically signiﬁcant (p = 0.148). Similarly, as highlighted in Fig. 4a there was no signiﬁcant dif- TABLE 1. Workload, heart rate and performance during laparoscopic and robotic suturing in self-paced and time pressure conditions. Self-paced Time pressure Error Error SURG-TLX DHR Task progres- score Leak vol- Knot tensile SURG-TLX Task progres- score Leak vol- Knot tensile score (au) (bpm) sion score (au) (mm) ume (mL) strength (N) Score (au) DHR sion score (au) (mm) ume (mL) strength (N) Laparoscopic 147.6 (52.2) 2 1.6 (3.2) 6.0 (1.0) 2.0 (2.0) 16.0 (2.4) 25.1 (74.4) 202.1 (46.2) 0.4 (4.4) 4.5 (2.0) 3.0 (3.6) 18.1 (3.5) 0.0 (15.4) suturing Robotic 115.9 (72.1) 1.7 (2.5) 6.0 (0.0) 1.9 (1.8) 15.5 (1.3) 21.2 (52.8) 183.9 (75.9) 1.0 (1.5) 5.0 (2.0) 2.1 (1.9) 18.3 (1.2) 1.4 (13.1) suturing SURG-TLX Surgical Task Load Index, DHR change in heart rate from rest to task, au arbitrary units, bpm beats per minute, mm millimetres, mL millilitres, N Newtons. SURG-TLX, DHR and Leak Volume data are mean values (SD). Task Progression Score, Error Score and Knot Tensile Strength data are median values (IQR). Robotic Surgery and Brain Function 1629 FIGURE 3. (a) Subjective workload (SURG-TLX) scores, heart rate, and technical performance in self-paced (yellow bars) and time pressure (green bars) conditions during robotic suturing. Error bars represent the 95% conﬁdence interval; (b) Comparison of the HbO response (DHbO ) in time pressure and self-paced conditions during robotic suturing. Channels exhibiting a smaller HbO 2 2 2 response in the time pressure compared to the self-paced condition are blue, whereas those exhibiting a smaller response in the self-paced condition are red. Channels in which there was a signiﬁcant difference (p < 0.05, Wilcoxon Signed Ranks test) in the HbO response between conditions are outlined (black circle). Reference points of the 10–20 system of optode placement are shown in yellow (right PFC: Fp2 = source 6, F8 = source 7, F2 = source 9, FC4 = source 10; left PFC: Fp1 = source 2, F7 = source 1, F1 = source 5, FC3 = source 4). Au arbitrary units, DHR change in heart rate from rest to task, bpm beats per minute, mm millimetres, mL millilitres, N Newtons. 1630 SINGH et al. FIGURE 4. (a) Subjective workload (SURG-TLX) scores, heart rate, and technical performance in laparoscopic (yellow bars) and robotic (green bars) suturing in the self-paced condition. Error bars represent the 95% conﬁdence interval; (b) Comparison of the HbO response (DHbO ) during laparoscopic and robotic suturing in the self-paced condition. Channels exhibiting a smaller HbO 2 2 2 response during robotic suturing compared to laparoscopic suturing are blue, whereas those exhibiting a smaller response during laparoscopic suturing are red. Channels in which there was a signiﬁcant difference (p < 0.05, Wilcoxon Signed Ranks test) in the HbO response between operative platforms are outlined (black circle). Reference points of the 10–20 system of optode placement are shown in yellow (right PFC: Fp2 = source 6, F8 = source 7, F2 = source 9, FC4 = source 10; left PFC: Fp1 = source 2, F7 = source 1, F1 = source 5, FC3 = source 4). Au arbitrary units, DHR change in heart rate from rest to task, bpm beats per minute, mm millimetres, mL millilitres, N Newtons. Robotic Surgery and Brain Function 1631 ference in DHR between laparoscopic and robotic changes in HbO in any channel in any subject during suturing (p = 0.092) (Table 1). laparoscopic suturing, and in one channel (channel 19 in subject 2) during robotic suturing. In the TP con- Technical Performance dition, signiﬁcant correlations were observed in two channels (channels 9 and 21 in subject 3) during There was no signiﬁcant diﬀerence in task progres- laparoscopic suturing, and in one channel (channel 4 in sion score (p = 0.296), error score (p = 0.391), leak subject 1) during robotic suturing (Supplementary volume (p = 0.640) or knot tensile strength Fig. 1). (p = 0.347) between laparoscopic and robotic suturing in the SP condition (see Fig. 4a and Table 1). DISCUSSION Cortical Haemodynamic Response The disparity in operator prefrontal cognitive de- In SP condition, DHbO was smaller during robotic mands between laparoscopy and robotic surgery was suturing compared to laparoscopic suturing in 21 explored for a surgical suturing task and was further channels, however these differences did not reach sig- contrasted by varying temporal urgency. The results niﬁcance (Fig. 4b). demonstrate the beneﬁts of the robotic platform during episodes of high mental demand. In the comparatively Laparoscopy vs. Robotic Surgery (Time Pressure) calm conditions of the self-paced paradigm, there was no discernible diﬀerence in technical performance Subjective Workload and Heart Rate between the two operative platforms. However, when SURG-TLX scores were lower during robotic the same skill had to be recreated with greater sense of suturing than laparoscopic suturing, but the diﬀerence temporal urgency, such as would be required to suture did not reach signiﬁcance threshold (p = 0.511). a bleeding vessel, the robotic platform enabled im- Similarly, as highlighted in Fig. 5a there was no sig- proved accuracy, reduced technical errors and accel- niﬁcant difference in DHR between laparoscopic and erated task progression. Critically, these performance robotic techniques (p = 0.743) (Table 1). improvements during the robotic task were coupled with interesting diﬀerences in the allocation of pre- Technical Performance frontal resources compared to laparoscopy. Both self-paced and time-pressure conditions were The robotic technique resulted in signiﬁcantly im- characterized by a general reduction in the magnitude of proved performance under TP with respect to task the prefrontal responses when operating with the ro- progression score (median (IQR): laparoscopic sutur- botic platform versus conventional laparoscopy. How- ing = 4.5 (2.0) vs. robotic suturing = 5.0 (2.0); ever, under temporal demand, the robotic platform was z = 2 2.107, p = 0.035) and error score (median associated with greater responses across the right (IQR): laparoscopic suturing = 3.0 (3.6) mm vs. ro- VLPFC, an area important for vigilance, resistance to botic suturing = 2.1 (1.9) mm; z = 2 2.488, 21,27 environmental distraction, and attentional control, p = 0.013). There were no signiﬁcant differences in as well as inappropriate motor response suppression. leak volume (p = 0.889) or knot tensile strength This would suggest improvements in technical perfor- (p = 0.719) between the robotic or laparoscopic plat- mance observed during robotic suturing under time forms (see Fig. 5a and Table 1). pressure compared to laparoscopic suturing reﬂect more focused attention and greater task engagement that the Cortical Haemodynamic Response robotic platform allows during stressful conditions, or In the TP Condition, greater prefrontal DHbO was alternatively that greater prefrontal resources are observed during robotic compared to laparoscopic required to operate with the robotic platform with which suturing in seven channels, particularly those located trainees were less accustomed. In this regard, our pre- in the right VLPFC (see Fig. 5b). vious work has shown that task novelty leads to a greater prefrontal response indicative of greater attentional demands which may explain the greater activation Correlation Analysis observed during robotic suturing. However, there was Across the all experimental conditions, only 4 out of no signiﬁcant difference in subjective workload (SURG- 768 channels demonstrated statistically signiﬁcantly TLX) score between platforms which suggests that, al- (p < 0.05) correlations (R > 0.8) between heart rate though participants were less experienced with the ro- and changes in HbO . In the SP condition, there were botic platform, they did not ﬁnd it more demanding than no signiﬁcant correlations between heart rate and the laparoscopic approach. 1632 SINGH et al. FIGURE 5. (a) Subjective workload (SURG-TLX) scores, heart rate, and technical performance in laparoscopic (yellow bars) and robotic (green bars) suturing in the time pressure condition. Error bars represent the 95% conﬁdence interval; (b) Comparison of the HbO response (DHbO ) during laparoscopic and robotic suturing in the time pressure condition. Channels exhibiting a smaller 2 2 HbO response during robotic suturing compared to laparoscopic suturing are blue, whereas those exhibiting a smaller response during laparoscopic suturing are red. Channels in which there was a signiﬁcant difference (p < 0.05, Wilcoxon Signed Ranks test) in the HbO response between operative platforms are outlined (black circle). Reference points of the 10–20 system of optode placement are shown in yellow (right PFC: Fp2 = source 6, F8 = source 7, F2 = source 9, FC4 = source 10; left PFC: Fp1 = source 2, F7 = source 1, F1 = source 5, FC3 = source 4). Au arbitrary units, DHR change in heart rate from rest to task, bpm beats per minute, mm millimetres, mL millilitres, N Newtons. Robotic Surgery and Brain Function 1633 Laparoscopic surgery has been shown to be more demand during laparoscopy was associated with a mentally demanding and/or stressful compared with relative decrease in DHbO . Similar prefrontal ‘deac- 31,37,40 robotic surgery. The current study is the ﬁrst to tivation’ responses during high workload states have show that this disparity is related to diminished pre- also been observed during mental arithmetic, simu- 13 16 frontal engagement and attentional control during la- lated piloting and military command tasks, and are paroscopy compared with the robotic approach. thought to be due to disruption of executive control 21,24 Whilst there have been no previous studies investigat- processes, distraction of participants with task-ir- ing this disparity from a neurocognitive perspective, relevant cues, or neuro-hormonal inhibition of pre- literature from outside medicine have similarly frontal synaptic activity. In contrast, robotic suturing demonstrated that mentally challenging tasks are under time pressure evoked a relative increase in pre- associated with attenuated prefrontal frontal cortical responses across lateral PFC regions. 1,13,16,26,34 responses. For example, in a fNIRS study by This implies that whilst there may be a mismatch Lin et al., subjects completed a walking task with or between workload and prefrontal haemodynamic without a concurrent n-back working memory task. responses during laparoscopic suturing under time The most mentally demanding conditions (i.e., dual pressure, there is improved alignment between esca- task performance) resulted in slower gait speed, greater lating mental demands and prefrontal activation dur- gait variability, and diminished prefrontal oxygena- ing robotic surgery. tion. In another fNIRS study prefrontal activation Here, widespread DHbO increases under time was assessed during a complex naval air warfare pressure, suggest greater brain-resource allocation to management task, in which task load was manipulated deal with the increasing mental demands, an observa- by changing the number of aircrafts to be managed (6, tion which is corroborated by others in the context of 5 6 12, 18, or 24) and by introducing an auditory memory working memory tasks, air trafﬁc control tasks, and 16 16 6 task. During the highest workload condition, mean warship command tasks. For example, Ayaz et al., cortical oxygenation decreased coupled with a deteri- used fNIRS to assess prefrontal activation in air trafﬁc oration in task performance. In line with ﬁndings controllers as they monitored an increasing number of from these fNIRS studies, Rowe et al., used fMRI to aircrafts, and observed that oxygenation in the lateral scan subjects performing a ﬁnger movement task dur- PFC increased in-line with escalating mental ing which their attention was directed either towards demands. Despite the alignment between temporal their actions (low cognitive demand) or towards a vi- pressure, subjective workload and prefrontal responses sual search task (high cognitive demand). During during robotic surgery it is interesting to note that attention-to-action, prefrontal activation increased PFC-related increase in attention and concentration compared with unattended performance of the same did not manifest in performance stabilization. It mat- 34 1 task. Finally, Al-Shargie et al., utilised multimodal tered little whether PFC deactivations (laparoscopy) or imaging (EEG and fNIRS) to measure prefrontal enhanced PFC activations (robotic) were observed, activity during a mental arithmetic task performed high temporal demand still led to technical perfor- under time pressure and with negative feedback. The mance degradation. However, as described above, ro- results showed that the stressful condition led to botic performance was superior to laparoscopic under diminished activation in the ventrolateral PFC, in line high temporal demands, possibly due to improved with current study ﬁndings. alignment between workload and prefrontal activa- As important as the direct between-platform com- tion, and enhanced attentional control and task parisons are the within-platform diﬀerences in work- engagement. It should be noted, that despite signiﬁcant load, performance and brain responses associated with improvement in task progression and error with ro- switching between self-paced and time pressure modes. botic suturing compared to the laparoscopic approach Regardless of the operative platform, time pressure led under time constraints, there were no differences in to greater perceived workload and signiﬁcant technical leak volume or knot tensile strength. This may have performance decline. Therefore, despite performance been a result of experimental design. For example, advantages of the robotic platform compared to la- subjects were not permitted to tie additional knots even paroscopy when temporal demands are high, perfor- if they felt the defect was not adequately closed, mance on the robot under pressure was substantially therefore any potential between-platform differences in inferior to self-paced performance. Although time leak volume would have been masked. Regarding knot pressure-related deterioration in technical performance tensile strength, the robotic platform lacks the force and escalation in subjective workload transcends feedback that the laparoscopic technique confers. operative platform, prefrontal responses associated Therefore, during robotic suturing, subjects would not with switching to a temporal demand were observed to have experienced the same tactile perception of knot be platform-dependent. Speciﬁcally, high temporal 1634 SINGH et al. tightness as they would have during laparoscopic knot ening HbO concentration change. Furthermore, tying. signiﬁcant lateral PFC haemodynamic change, may be High workload states, such as temporal demands, confounded by temporalis muscle activity. Future could precipitate a stress-induced increase in heart assessment of the cytochrome c oxidase response, a rate, prompting spurious haemodynamic responses. more brain-speciﬁc biomarker of cerebral metabolism, 12,19 Indeed, associations between heart rate and right PFC may overcome these limitations. A pre-hoc power 7,39 activity have been reported. However, mental calculation was not possible as data from pre-existing arithmetic was used in previous studies to increase studies was inadequate for a sample size estima- 7,39 5,6,8,13,16 cognitive demands which may not extrapolate to tion, which others have similarly not 5,6,8,13,16 complex surgical tasks. Moreover, there were no sig- attempted. Only eight subjects participated niﬁcant differences in DHR between conditions or with others citing time required (approximately 3 h per between platforms in the current study, and signiﬁcant subject) as prohibitive. This notwithstanding, numer- per-subject correlations between heart rate and chan- ous dependant variables were measured in each sub- ges in HbO were not endemic, suggesting the results ject—(Hb species concentrations, heart rate, etc.) and reﬂect genuine differences in cognitive processing. hence the multidimensional dataset comprised thou- Interestingly, these correlations were subject-speciﬁc sands of data points which increased the validity of the rather than related to expertise or platform speciﬁc. statistical analysis. Whilst the controlled laboratory This is not the ﬁrst paper to describe to the relative environment and use of a bench-top laparoscopic merits of robotic surgery compared to laparoscopy. trainer may not mirror the operating room, between- Studies have shown that, compared to the laparoscopic condition differences in SURG-TLX scores imply the approach, robotic surgery improves manual dexter- 2 min restriction adequately recreated a sense of ur- 10 31,37 ity, reduces technical errors, and shortens learn- gency and increased mental effort. Moreover, the face 10,37 9,25,33,36 ing curves in novices. Furthermore, robotic surgery validity of low-cost box trainers is proven. 31,37,40 has been shown to reduce cognitive burden and 31,40 dampen the physiological stress response. Whilst ELECTRONIC SUPPLEMENTARY MATERIAL these studies all provide evidence to support the beneﬁts of robotic platforms, the current is the ﬁrst to demon- The online version of this article (https://doi.org/10. strate the relative beneﬁts in terms of improved task- 1007/s10439-018-2049-z) contains supplementary related attention and concentration using direct and material, which is available to authorized users. objective assessment of surgeons’ brain function during episodes of high mental workload. ACKNOWLEDGMENTS CONCLUSION We would like to acknowledge the infrastructural support of the Imperial NIHR Biomedical Research During episodes of high mental workload evoked by Centre and Imperial Cancer Research UK Centre. temporal demands the robotic platform leads to im- proved technical performance and greater alignment between temporal demands, subjective workload and CONFLICT OF INTEREST prefrontal activation as compared to laparoscopy. Further work should seek to develop biofeedback No beneﬁts in any form have been or will be re- interventions to modulate neural activation during ceived from a commercial party related directly or temporal stress and high workload states with the aim indirectly to the subject of this manuscript. of improving attention and concentration, enhancing performance and improving patient safety. FUNDING The work was funded through the support of the Imperial NIHR BRC and Imperial Cancer Research LIMITATIONS UK Centre. Changes in respiration can lead to ﬂuctuations in partial pressure of CO (pCO ), altering cerebral blood 2 2 OPEN ACCESS 17,43 ﬂow through changes in cerebral vascular tone. This article is distributed under the terms of the Although respiration was not measured, time-pressure Creative Commons Attribution 4.0 International Li- may precipitate stress-related hyperventilation, de- cense (http://creativecommons.org/licenses/by/4.0/), creases in pCO and cerebral vasoconstriction, damp- 2 Robotic Surgery and Brain Function 1635 which permits unrestricted use, distribution, and re- to detect mental overload. Behav. Brain Res. 259:16–23, production in any medium, provided you give appro- Hubert, N., M. Gilles, K. Desbrosses, J. P. Meyer, J. Fel- priate credit to the original author(s) and the source, blinger, and J. Hubert. Ergonomic assessment of the sur- provide a link to the Creative Commons license, and geon’s physical workload during standard and robotic indicate if changes were made. assisted laparoscopic procedures. Int. J. Med. Robot 9(2):142–147, 2013. Huppert, T. J., S. G. Diamond, M. A. Franceschini, and D. A. Boas. HomER: a review of time-series analysis methods REFERENCES for near-infrared spectroscopy of the brain. Appl. Opt. 48(10):280–298, 2009. 1 16 Al-Shargie, F., M. Kiguchi, N. Badruddin, S. C. Dass, A. Izzetoglu, K., S. Bunce, B. Onaral, K. Pourrezaei, and B. F. Hani, and T. B. Tang. Mental stress assessment using Chance. functional optical brain imaging using near-in- simultaneous measurement of EEG and fNIRS. Biomed. frared during cognitive tasks. Int. J. Human Comput. Opt. Express 7(10):3882–3898, 2016. Interact. 17(2):211–227, 2004. 2 17 Arnsten, A. F. T. Stress signalling pathways that impair Jordan, J., J. R. Shannon, A. Diedrich, B. Black, F. Costa, prefrontal cortex structure and function. Nat. Rev. Neu- D. Robertson, and I. Biaggioni. Interaction of carbon rosci. 10(6):410–422, 2009. dioxide and sympathetic nervous system activity in the Aron, A. R., T. W. Robbins, and R. A. Poldrack. Inhibi- regulation of cerebral perfusion in humans. Hypertension tion and the right inferior frontal cortex. Trends Cogn. Sci. 36(3):383–388, 2000. 8(4):170–177, 2004. Jurcak, V., D. Tsuzuki, and I. Dan. 10/20, 10/10, and 10/5 Arora, S., N. Sevdalis, D. Nestel, M. Woloshynowych, A. systems revisited: their validity as relative head-surface- Darzi, and R. Kneebone. The impact of stress on surgical based positioning systems. Neuroimage 34(4):1600–1611, performance: a systematic review of the literature. Surgery 147(3):318–330, 2010. Kolyva, C., A. Ghosh, I. Tachtsidis, D. Highton, C. E. Ayaz, H., M. Izzetoglu, S. Bunce, T. Heiman-Patterson Cooper, M. Smith, and C. E. Elwell. Cytochrome c oxidase and B. Onaral. Detecting cognitive activity related hemo- response to changes in cerebral oxygen delivery in the adult dynamic signal for brain computer interface using func- brain shows higher brain-speciﬁcity than haemoglobin. tional near infrared spectroscopy. In: 3rd International Neuroimage 85(Pt 1):234–244, 2014. IEEE/EMBS Conference on Neural Engineering, 2007. pp. Lanfranco, A. R., A. E. Castellanos, J. P. Desai, and W. C. 342–345. Meyers. Robotic surgery: a current perspective. Ann. Surg. Ayaz, H., P. A. Shewokis, S. Bunce, K. Izzetoglu, B. 239(1):14–21, 2004. Willems, and B. Onaral. Optical brain monitoring for Langner, R., and S. B. Eickhoff. Sustaining attention to operator training and mental workload assessment. Neu- simple tasks: A meta-analytic review of the neural mecha- roimage 59(1):36–47, 2012. nisms of vigilant attention. Psychol. Bull. 139(4):870–900, Brugnera, A., C. Zarbo, R. Adorni, G. A. Tasca, M. 2013. Rabboni, E. Bondi, A. Compare, and K. Sakatani. Corti- Leff, D. R., C. E. Elwell, F. Orihuela-Espina, L. Atallah, cal and cardiovascular responses to acute stressors and D. T. Delpy, A. W. Darzi, and G. Z. Yang. Changes in their relations with psychological distress. Int. J. Psy- prefrontal cortical behaviour depend upon familiarity on a chophysiol. 114:38–46, 2017. bimanual co-ordination task: an fNIRS study. Neuroimage Bunce, S. C., K. Izzetoglu, H. Ayaz, P. Shewokis, M. 39(2):805–813, 2008. Izzetoglu, K. Pourrezaei and B. Onaral. Implementation of Leff, D. R., F. Orihuela-Espina, T. Athanasiou, V. Kar- fNIRS for monitoring levels of expertise and mental imyan, C. Elwell, J. Wong, G. Z. Yang, and A. W. Darzi. workload. International Conference on Foundations of ‘‘Circadian cortical compensation’’: a longitudinal study of Augmented Cognition. Orlando, Florida USA: Springer brain function during technical and cognitive skills in Berlin Heidelberg, 2011. pp. 13–22. acutely sleep-deprived surgical residents. Ann. Surg. Campo, R., C. Reising, Y. Van Belle, J. Nassif, P. 252(6):1082–1090, 2010. O’Donovan, and C. R. Molinas. A valid model for testing Leff, D. R., G. Yongue, I. Vlaev, F. Orihuela-Espina, D. and training laparoscopic psychomotor skills. Gynecol. James, M. J. Taylor, T. Athanasiou, R. Dolan, G. Z. Yang, Surg. 7(2):133–141, 2010. and A. Darzi. ‘‘Contemplating the Next Maneuver’’: Chandra, V., D. Nehra, R. Parent, R. Woo, R. Reyes, T. functional neuroimaging reveals intraoperative decision- Hernandez-Boussard, and S. Dutta. A comparison of making strategies. Ann. Surg. 265(2):320–330, 2017. laparoscopic and robotic assisted suturing performance by Li, M. M., and J. George. A systematic review of low-cost experts and novices. Surgery 147(6):830–839, 2010. laparoscopic simulators. Surg. Endosc. 31(1):38–48, 2017. Cope, M., D. T. Delpy, E. O. Reynolds, S. Wray, J. Wyatt, Lin, M.-I. B., and K.-H. Lin. Walking while performing and P. van der Zee. Methods of quantitating cerebral near working memory tasks changes the prefrontal cortex infrared spectroscopy data. Adv. Exp. Med. Biol. 222:183– hemodynamic activations and gait kinematics. Front. Be- 189, 1988. hav. Neurosci. 10:92, 2016. 12 27 de Roever, I., G. Bale, R. J. Cooper, and I. Tachtsidis. Love, T., F. Haist, J. Nicol, and D. Swinney. A functional Functional NIRS measurement of cytochrome-C-oxidase neuroimaging investigation of the roles of structural com- demonstrates a more brain-speciﬁc marker of frontal lobe plexity and task-demand during auditory sentence pro- activation compared to the haemoglobins. Adv. Exp. Med. cessing. Cortex 42(4):577–590, 2006. Biol. 977:141–147, 2017. Miller, E. K., and J. D. Cohen. An integrative theory of Durantin, G., J. F. Gagnon, S. Tremblay, and F. Dehais. prefrontal cortex function. Annu. Rev. Neurosci. 24:167– Using near infrared spectroscopy and heart rate variability 202, 2001. 1636 SINGH et al. Modi, H. N., H. Singh, F. Orihuela-Espina, T. Athanasiou, training using inexpensive box trainers: which exercises to F. Fiorentino, G. Z. Yang, A. Darzi, and D. R. Leff. choose when constructing a validated training course. Temporal stress in the operating room: brain engagement BJOG 118(13):1576–1584, 2011. promotes ‘‘Coping’’ and disengagement prompts ‘‘Chok- Stefanidis, D., F. Wang, J. R. Korndorffer, J. B. Dunne, ing’’. Ann. Surg. 267:683–691, 2017. and D. J. Scott. Robotic assistance improves intracorporeal Modi, H. N., H. Singh, G. Z. Yang, A. Darzi, and D. R. suturing performance and safety in the operating room Leff. A decade of imaging surgeons’ brain function (part while decreasing operator workload. Surg. Endosc. II): a systematic review of applications for technical and 24(2):377–382, 2010. nontechnical skills assessment. Surgery 162(5):1130–1139, Tachtsidis, I., and F. Scholkmann. False positives and false 2017. negatives in functional near-infrared spectroscopy: issues, Moore, L. J., M. R. Wilson, J. S. McGrath, E. Waine, R. S. challenges, and the way forward. Neurophotonics W. Masters, and S. J. Vine. Surgeons’ display reduced 3(3):031405, 2016. mental effort and workload while performing robotically Tanida, M., K. Sakatani, R. Takano, and K. Tagai. assisted surgical tasks, when compared to conventional Relation between asymmetry of prefrontal cortex activities laparoscopy. Surg. Endosc. 29(9):2553–2560, 2015. and the autonomic nervous system during a mental arith- Orihuela-Espina, F., D. R. Leff, D. R. C. James, A. W. metic task: near infrared spectroscopy study. Neurosci. Darzi, and G. Z. Yang. Imperial College near infrared Lett. 369(1):69–74, 2004. spectroscopy neuroimaging analysis framework. Neu- van der Schatte Olivier, R. H., C. D. P. van‘t Hullenaar, J. rophotonics 5(1):011011, 2018. P. Ruurda, and I. A. M. J. Broeders. Ergonomics, user Palter, V. N., N. Orzech, R. Aggarwal, A. Okrainec, and T. comfort, and performance in standard and robot-assisted P. Grantcharov. Resident perceptions of advanced laparoscopic surgery. Surg. Endosc. 23(6):1365, 2008. laparoscopic skills training. Surg. Endosc. 24(11):2830– Wilson, M. R., J. M. Poolton, N. Malhotra, K. Ngo, E. 2834, 2010. Bright, and R. S. Masters. Development and validation of a Rowe, J., K. Friston, R. Frackowiak, and R. Passingham. surgical workload measure: the surgery task load index Attention to action: speciﬁc modulation of corticocortical (SURG-TLX). World J. Surg. 35(9):1961–1969, 2011. interactions in humans. Neuroimage 17(2):988–998, 2002. Yang, G.-Z., J. Cambias, K. Cleary, E. Daimler, J. Drake, Schecklmann, M., A. Mann, B. Langguth, A. C. Ehlis, A. P. E. Dupont, N. Hata, P. Kazanzides, S. Martel, R. V. J. Fallgatter, and F. B. Haeussinger. The temporal muscle Patel, V. J. Santos, and R. H. Taylor. Medical of the head can cause artifacts in optical imaging studies robotics—Regulatory, ethical, and legal considerations for with functional near-infrared spectroscopy. Front. Hum. increasing levels of autonomy. Sci. Robot. 2(4):8638, 2017. Neurosci. 11:456, 2017. Yoon, S., M. Zuccarello, and R. M. Rapoport. pCO(2) and Schreuder, H. W., C. B. van den Berg, E. J. Hazebroek, R. pH regulation of cerebral blood ﬂow. Front. Physiol. 3:365, H. Verheijen, and M. P. Schijven. Laparoscopic skills 2012.
Annals of Biomedical Engineering – Springer Journals
Published: Jun 4, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera