Abstract
Fluorescence videography is a promising technique for assessing bowel perfusion. Fluorescence-based enhanced reality (FLER) is a novel concept, in which a dynamic perfusion cartogram, generated by computer analysis, is superimposed on to real-time laparoscopic images. The aim of this experimental study was to assess the accuracy of FLER in detecting differences in perfusion in a small bowel resection–anastomosis model.
A small bowel ischaemic segment was created laparoscopically in 13 pigs. Animals were allocated to having anastomoses performed at either low perfusion (25 per cent; n = 7) or high perfusion (75 per cent; n = 6), as determined by FLER analysis. Capillary lactate levels were measured in blood samples obtained by serosal puncturing in the ischaemic area, resection lines and vascularized areas. Pathological inflammation scoring of the anastomosis was carried out.
Lactate levels in the ischaemic area (mean(s.d.) 5·6(2·8) mmol/l) were higher than those in resection lines at 25 per cent perfusion (3·7(1·7) mmol/l; P = 0·010) and 75 per cent perfusion (2·9(1·3) mmol/l; P < 0·001), and higher than levels in vascular zones (2·5(1·0) mmol/l; P < 0·001). Lactate levels in resection lines with 75 per cent perfusion were lower than those in lines with 25 per cent perfusion (P < 0·001), and similar to those in vascular zones (P = 0·188). Levels at resection lines with 25 per cent perfusion were higher than those in vascular zones (P = 0·001). Mean(s.d.) global inflammation scores were higher in the 25 per cent perfusion group compared with the 75 per cent perfusion group for mucosa/submucosa (2·1(0·4) versus 1·2(0·4); P = 0·003) and serosa (1·8(0·4) versus 0·8(0·8); P = 0·014). A ratio of preanastomotic lactate levels in the ischaemic area relative to the resection lines of 2 or less was predictive of a more severe inflammation score.
In an experimental model, FLER appeared accurate in discriminating bowel perfusion levels.
Surgical relevance
Clinical assessment has limited accuracy in evaluating bowel perfusion before anastomosis. Fluorescence videography estimates intestinal perfusion based on the fluorescence intensity of injected fluorophores, which is proportional to bowel vascularization. However, evaluation of fluorescence intensity remains a static and subjective measure.
Fluorescence-based enhanced reality (FLER) is a dynamic fluorescence videography technique integrating near-infrared endoscopy and specific software. The software generates a virtual perfusion cartogram based on time to peak fluorescence, which can be superimposed on to real-time laparoscopic images. This experimental study demonstrates the accuracy of FLER in detecting differences in bowel perfusion in a survival model of laparoscopic small bowel resection–anastomosis, based on biochemical and histopathological data.
It is concluded that real-time imaging of bowel perfusion is easy to use and accurate, and should be translated into clinical use.
Introduction
Adequate bowel perfusion is a key determinant of successful healing1. An accurate and objective intraoperative method for evaluating intestinal microperfusion is needed as clinical assessment has limited accuracy2. A promising tool for image-guided intraoperative assessment of the future anastomotic site in minimally invasive procedures is fluorescence videography. It employs a near-infrared endoscope capable of detecting the signal emitted by a fluorescent dye, frequently indocyanine green (ICG), which is administered by intravenous injection. Fluorescence intensity is proportional to the amount of fluorescent dye diffused in the tissue and is consequently a marker of tissue perfusion3. Estimation of fluorescence intensity is a rapid, user-friendly approach, which is completely integrated into the surgical workflow, and could potentially become part of routine practice. To date, only a few investigators4–7 have undertaken fluorescence-based evaluation of bowel perfusion in the clinical setting, using different commercially available devices designed for minimally invasive surgery.
In previous experiments, perfusion was evaluated in a static fashion, based merely on relative fluorescence intensity, without considering the diffusion of fluorophores over time. The dye can reach the boundaries of ischaemic areas by capillary flow diffusion over time, and the perfused zone may be overestimated. Additionally, fluorescent intensity is highly dependent on the distance between the light source and the target, and requires the use of a standard reference calibration aid (a squared spot that yields a constant signal when illuminated by the near-infrared light)8–10 introduced in the abdomen.
The research and development department at the Institute for Research Against Cancer of the Digestive System (IRCAD) has developed dedicated image analyser software (ER-PERFUSION) to obtain a dynamic perfusion cartogram based on fluorescence time-to-peak measurements8–10.
The first step in the development of fluorescence-based enhanced reality (FLER) was to establish proof of concept, in a 1-h ischaemia non-survival model10, in which the software was set to identify a drop in perfusion of 50 per cent; this was correlated with capillary lactate levels, impairment of mitochondrial respiration, and an early ischaemia signature defined by high-resolution magic-angle-spinning 1H nuclear magnetic resonance analysis.
In a second non-survival study, the aim was to compare FLER performance (again set at 50 per cent perfusion) and clinical judgement in recognizing ischaemic areas. Clinical assessments were done by experienced surgeons after 2, 4 and 6 h of ischaemia9. Based on the same surrogate markers, FLER was superior to clinical evaluation, particularly after 4 h of ischaemia. The next step comprised a further validation study, based on the simultaneous evaluation of the serosal side (by FLER, set at 50 per cent perfusion) and of mucosal endomicroscopic changes using a probe-based endoscopic confocal system, in a 1-h sigmoid ischaemia model. Again, FLER was able to identify the areas with early signs of morphological and cellular damage more precisely than clinical estimation8.
The aim of the present study was to assess the FLER technique in an experimental survival study of bowel ischaemia. Predetermined degrees of perfusion were imposed by the software, with the aim of evaluating the accuracy of the technique, based on chemical surrogate markers of perfusion, such as capillary lactate levels and mitochondrial chain respiratory rate, as well as on clinical and morphological outcomes.
Methods
This experimental study (protocol no. 38.2012.01.039) received full approval from the local Ethics Committee on Animal Experimentation. All animals used in the experimental laboratory were managed according to Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines11, French laws for animal use and care, and directives from the European Community Council (2010/63/EU). Pigs were fasted for 24 h before surgery, with free access to water. Premedication was administered 1 h before surgery by intramuscular injection of ketamine (20 mg/kg) and azaperone (2 mg/kg) (Stressnil; Janssen-Cilag, Beerse, Belgium). Induction was achieved with intravenous propofol (3 mg/kg) combined with pancuronium (0·2 mg/kg). Anaesthesia was maintained with 2 per cent isoflurane. Warming blankets were placed below the animals to prevent hypothermia. At the end of the protocol, animals were killed by intravenous injection of a lethal dose of potassium chloride. Thirteen Large White pigs, mean(s.d.) weight at inclusion 27·1(3·1) kg (3 male, 10 female), were used in this 10-day survival study.
Fluorescence-based enhanced reality and construction of perfusion cartogram
FLER uses a fluorescence videography system, integrating a near-infrared endoscope (D-Light P; Karl Storz Endoscope, Tuttlingen, Germany) that is able to detect the fluorescence signal emitted by ICG, and ER-PERFUSION software, which generates a dynamic perfusion cartogram10. The software runs on a laptop connected to the laparoscopic unit. Perfusion cartograms can be superimposed on to real-time laparoscopic images to obtain an enhanced reality effect. The dynamic perfusion cartogram of the bowel is created by averaging fluorescence signals over a 20–40-s video at a speed of 5–25 frames per second, and by assigning a colour code based on the time required to reach the maximum intensity of each pixel. Each single pixel of a perfusion cartogram is a dynamic image (two dimensions plus diffusion time) representing the average of multiple images (between 100 and 200) (Fig. 1). The perfusion cartography information is stored in the software memory, and allows visualization of the level of perfusion by navigating with the computer mouse over the surface of the bowel in white light as it was imaged at the moment of obtaining the fluorescence signal in near-infrared mode. The perfusion cartogram is obtained in 2–3 min (40 s for fluorescence signal development and 1–2 min to place the regions of interest (ROIs) on the bowel). Additional time, less than 1 min, is required to register the synthetic image on real-time images.
Dynamic perfusion cartography. The dynamic perfusion cartogram was created by averaging fluorescence signals over a 20–40-s video at a speed of 5–25 frames per second and by attributing a colour code based on the slope intensity/time of each pixel. Each single pixel of the cartogram results from the average of multiple subsequent images
This analysis is independent of distance to target, as it is based on the steepness of the intensity/time slope, which is not affected by distance. This feature also allows correction of the gradient of intensity between the centre of the image and the edges, because the slope remains constant in the observed field. Time to reach fluorescence peak correlates inversely with perfusion level.
Surgical procedure
Pigs were allocated to one of two groups according to differences in perfusion (25 per cent, n = 7; 75 per cent, n = 6), relative to a 100 per cent vascular area of the bowel, as assessed by FLER (Video S1, supporting information). Each animal acted as its own control. Three ports were placed and a 12-mmHg pneumoperitoneum was established. An ischaemic segment was created in a small bowel loop and 0·5 mg/kg ICG was injected 2 h later. The D-Light P laparoscope was then switched to near-infrared mode. ER-PERFUSION software was used to analyse the fluorescence signal and generated virtual perfusion cartograms based on time to peak fluorescence. Resection lines were displayed by superimposing virtual perfusion cartogram images on to the laparoscopic screen, thus obtaining enhanced reality. The cartogram was modulated to show the lines where perfusion dropped to 75 per cent and to 25 per cent, compared with vascular areas, in each loop.
ROIs were labelled as shown in Fig. 2. The identified levels of perfusion (25 or 75 per cent) were marked differently according to the allocated groups. In the 25 per cent perfusion group, surgical clips marked the 75 per cent level following FLER indications, and the 25 per cent level was marked with a surgical pen. Conversely, in the 75 per cent group, the 25 per cent perfusion level was marked using clips, and the 75 per cent level by means of a surgical pen. After marking, a minilaparotomy was performed and the loop was removed. The ischaemic area was resected following pen lines in all animals, and a handsewn end-to-end bowel anastomosis was performed within 30 min of FLER analysis. In the 25 per cent group, clips marking the 75 per cent perfusion level were removed from the bowel after measurement of capillary lactate levels.
Regions of interest (ROIs) identified by fluorescence-based enhanced reality analysis were conventionally labelled as follows: ROI 1, ischaemic; ROI 2a (25 and 75 per cent perfusion), left resection line; ROI 2b (25 and 75 per cent perfusion), right resection line; ROI 3a, left control area; ROI 3b, right control area
Measurement of capillary lactate levels
Preanastomotic capillary lactate levels were measured in all animals during the procedure using an EDGE® lactate analyser (ApexBio, Taipei, Taiwan, China) in the ischaemic area (ROI 1), resection lines ROI 2a (left) and ROI 2b (right), according to both 25 and 75 per cent perfusion in each loop, and in two vascularized areas (ROI 3a, left; ROI 3b, right) of the same loop to serve as control.
At the second look (postoperative day (POD) 7–12), postanastomotic capillary lactate levels were measured at the anastomotic site resulting from the anastomosis of 2a and 2b, and from two perianastomotic areas (3a and 3b), 2 cm apart from the anastomosis line.
Mitochondrial respiratory rate
Surgical biopsies were taken at POD 0 in the ischaemic area (ROI 1) and from resection lines ROI 2a + 2b as identified by FLER (according to the predetermined level of perfusion, 25 or 75 per cent, corresponding to the anastomotic level) to measure the mitochondrial respiratory rate. At the second look, samples were taken from the anastomosis and from two perianastomotic areas (3a and 3b), 2 cm apart from the anastomotic line (on the left and right respectively). The biopsy site on the anastomosis was identified by a suture. Samples were placed in a 2-ml water-jacketed oxygraphic cell (Oxygraph-2 k©; Oroboros Instruments, Innsbrück, Austria) equipped with a Clark electrode. After determination of basal oxygen consumption (V0), maximum small bowel tissue respiratory rates (Vmax) were measured on the same samples at 37°C in the presence of a saturating amount of ADP as phosphate acceptor. The Oroboros® oxygraphic system can test only two samples at a time and the analysis lasted 1 h. Owing to these technical limitations, ROI 1 was tested against 2a and 2b (resection lines) during the procedure but not against a definitely vascularized area. At the 10-day evaluation, the mitochondrial respiratory rate at the anastomosis was compared with two perianastomotic areas, representing vascular areas.
Exploratory laparotomy was performed to rule out anastomotic complications. The presence and extent of adhesions were scored following a classification adapted from Nair and colleagues12.
Histopathological assessment
The specimens were fixed in 4 per cent buffered formalin for at least 24 h. Sections 4 µm thick were cut from paraffin-embedded tissues, and stained with haematoxylin and eosin. Six sections per anastomosis were analysed. Assessments were made for the presence of fibroblasts, polynuclear neutrophils, lymphocytes and macrophages. A semiquantitative scoring system was used by an independent pathologist to evaluate the extent and severity of inflammation.
Statistical analysis
Data are presented as mean(s.d.) unless indicated otherwise. ANOVA followed by Dunnett's multiple comparison test was used for comparisons of lactate data from ischaemic zones, future resection lines and vascularized areas. Fisher's exact test was used to calculate P values for categorical variables. Student's t test was used for analysis of histopathological scores and mitochondrial respiratory rate. P < 0·050 was considered statistically significant. Statistical analysis was carried out using GraphPad Prism® software (GraphPad Software, La Jolla, California, USA).
Results
One pig was killed on POD 1 owing to flaccid palsy caused by bilateral cortical stroke, as demonstrated by brain MRI. This complication resulted from a problem with endotracheal intubation, which was noticed at the end of the procedure. Data from this animal were excluded from the analysis. There were no technical intraoperative complications. The survival period ranged from 7 to 12 days. On POD 6, one pig from the 25 per cent perfusion group was prostrate, with a distended abdomen and ileus. Exploratory laparotomy on POD 7 revealed a frank anastomotic leak with faecal peritonitis, and the animal was killed. One pig in the 75 per cent perfusion group developed ileus on POD 2. Exploratory laparoscopy was negative for leaks or any other intra-abdominal problems, and the animal survived without any further complication. Mean weight gain was 2·4(1·5) kg.
Analysis of resection lines
FLER analysis was performed effectively in all pigs in a maximum of 10 min, for both 25 and 75 per cent marking. The mean distance between resection lines identified by FLER at 25 and 75 per cent perfusion was 1·5(1·2) cm.
Capillary lactate levels
Before anastomosis
Mean capillary lactate levels in the ischaemic area (ROI 1) were 5·6(2·8) mmol/l, and were significantly higher than levels in resection lines 2a + 2b with 25 per cent perfusion (3·7(1·7) mmol/l; P = 0·010) or 75 per cent perfusion (2·9(1·3) mmol/l; P < 0·001), and higher than values in vascular zones 3a + 3b (2·5(1·0) mmol/l; P < 0·001). Lactate levels were significantly lower in resection lines 2a + 2b with 75 per cent perfusion than in those with 25 per cent perfusion (P < 0·001), and were similar to levels in vascular zones (P = 0·188). Values in resection lines 2a + 2b with 25 per cent perfusion were significantly higher than those in vascular zones (P = 0·001) (Fig. 3a).
Mean(s.d.) capillary lactate levels a before and b after anastomosis in regions of interest (ROIs) with different perfusion levels (25 or 75 per cent) identified by fluorescence-based enhanced reality. *P < 0·050 (ANOVA followed by Dunnett's multiple comparisons). c Relationship between preanastomotic lactate levels (ratio of ROI 1/2a + 2b) and histological inflammation score
After anastomosis
The mean capillary lactate level at the anastomosis was 4·0(1·9) and 2·9(1·4) mmol/l in the 25 and 75 per cent perfusion groups respectively (P = 0·326). Levels in the perianastomotic areas 3a + 3b were significantly higher in the 25 per cent group (3·9(1·8) versus 2·2(1·2) mmol/l; P = 0·019) (Fig. 3b). The ratio between lactate levels at the anastomosis and those in vascularized areas was significantly higher in the 25 per cent perfusion group than in the 75 per cent perfusion group (0·9(0·1) versus 0·7(0·2); P = 0·019).
Mitochondrial respiratory rate
Before anastomosis
Mean V0 in ROI 1 was 40·2(12·7) pmol oxygen per s per mg dry weight, and was significantly lower than that for resection lines 2a + 2b at 75 per cent perfusion (57·7(22·8) pmol/s; P = 0·036), whereas there was no difference between V0 at ROI 1 and V0 measured at 2a + 2b with 25 per cent perfusion (50·1(20·3) pmol/s; P = 0·231) (Fig. 4a). Vmax in ROI 1 (68·4(39·4) pmol oxygen per s per mg dry weight) was significantly lower than that in resection lines 2a + 2b at 25 per cent perfusion according to FLER (111·6(51·0) pmol/s; P = 0·022) and at 75 per cent perfusion (107·6(48·6) pmol/s; P = 0·036) (Fig. 4b). No statistically significant difference was found in V0 and Vmax between resection lines identified by FLER at 25 and 75 per cent perfusion.
Mitochondrial respiratory rate. Mean(s.d.) a basal oxygen consumption (V0) and b maximum respiratory rate (Vmax) before and after anastomosis in regions of interest (ROI) with perfusion levels of 25 or 75 per cent identified by fluorescence-based enhanced reality. *P < 0·050 (Student's t test)
After anastomosis
Mean V0 at the anastomosis (formed by the connection of 2a and 2b) was 50·4(24·2) and 62·1(17·6) pmol oxygen per s per mg dry weight in the 25 and 75 per cent perfusion groups respectively (P = 0·216). There was a significant difference between V0 in ROI 1 (before anastomosis) and in biopsies sampled after the survival period in control areas 3a + 3b (63·2(33·8) and 76·8(19·5) pmol/s for 25 and 75 per cent perfusion respectively; P = 0·044 and P < 0·001). Mean Vmax at the anastomosis was similar in both groups (87·9(36·1) and 86·7(34·1) pmol/s in 25 and 75 per cent perfusion groups respectively). Vmax in vascular perianastomotic areas (3a + 3b) was significantly higher than at the anastomosis in the 75 per cent group (P = 0·029).
Exploratory laparotomy
The mean adhesion score was 2·6(1·3) and 1·4(1·5) in the 25 and 75 per cent perfusion groups respectively (P = 0·056). In two pigs in the 25 per cent group, there were clear signs of anastomotic leakage with perianastomotic abscesses and faecal peritonitis. There was no anastomotic leakage in the 75 per cent group. One pig presented with multiple and diffuse adhesions but the anastomotic site was completely clear of adhesions (Table S1, supporting information).
Histopathological evaluation of the anastomosis
The results of histopathological evaluation are summarized in Table S2 (supporting information). At macroscopic evaluation of the anastomosis, the mean extent of ulceration was 2·7(3·0) and 1·4(1·1) mm in the 25 and 75 per cent perfusion groups respectively (P = 0·374).
Mean global inflammation scores were significantly higher in the 25 per cent perfusion group compared with the 75 per cent perfusion group for the mucosa/submucosa (2·1(0·4) versus 1·2(0·4); P = 0·003) and the serosa (1·8(0·4) versus 0·8(0·8); P = 0·014). A ratio of preanastomotic lactate levels in the ischaemic area (ROI 1) and resection lines (ROI 2a + 2b, future anastomotic segments) of 2 or less was predictive of a more severe inflammation score (Fig. 3c).
Discussion
The real-time FLER system allows quantitative assessment of bowel perfusion to determine the optimal anastomotic site. The present study represents the last preclinical step validating the FLER system in a survival animal model before progressing to clinical assessment.
The quantitative definition of critical preanastomotic perfusion is not known. It was felt that a subcritical model of perfusion would increase the likelihood of anastomotic complications and reduce the number of experimental subjects. When designing the protocol, balancing the weight of ethical and statistical considerations, a critical perfusion of 25 per cent of the maximum was considered acceptable given the possibility of reducing the required sample size. Lower rates of perfusion were considered to present too great a theoretical risk of substantial suffering to the animals, with no additional scientific benefit.
FLER analysis can be used to evaluate perfusion at any location in the gastrointestinal tract. This model should be considered as a simulation of anastomotic perfusion and not as a small bowel mesenteric ischaemia model, in which the whole ischaemic segment would be removed, without paying too much attention to resection lines. In colorectal resections, any single centimetre of the anastomotic line can have an impact on tension and on the healing process, and a system able to display the exact perfusion value might have significant clinical impact.
Local capillary lactate levels reflect tissue oxygenation more closely than systemic levels13. A clinical trial (NCT01634815) is currently under way to evaluate chronic critical leg ischaemia before and after revascularization using capillary lactate levels, measured by means of a strip-based lactate analyser that requires a single drop of blood.
The impact of preanastomotic capillary lactate levels on healing digestive anastomoses has not been evaluated previously. As expected, FLER analysis found a significant difference in local lactate levels between areas perfused at 25 and 75 per cent. Although not statistically significant, owing to the limited sample size, anastomotic complications (n = 2) occurred only in the 25 per cent perfusion group and postoperative adhesion scores were also higher in this group, along with higher mean preanastomotic local lactate levels. A ratio of 2 or less between lactate levels in the ischaemic area and the resection lines (2a + 2b) was predictive of higher inflammation scores on microscopic analysis. From these limited data, 25 per cent seems to represent a subcritical level of perfusion, at least in this porcine model, which certainly impairs healing but does not necessarily generate surgical complications.
The oxygraphic evaluation of mitochondrial chain respiratory rate is a robust tool used in ischaemia–reperfusion studies14, allowing quantitative functional estimation of various mitochondrial enzyme complexes involved in oxidative phosphorylation. In previous experiments8–10, early significant respiratory chain impairments in ischaemic zones compared with vascularized zones were expressed as a significant reduction in basal (V0) and maximum (Vmax) respiratory rate. In the present study the precision of FLER was evaluated by determining preanastomotic values of V0 and Vmax in the ischaemic area and at regions with 25 or 75 per cent perfusion. As expected, before anastomosis, the ischaemic small bowel demonstrated reduced mitochondrial respiration. Importantly, V0 at the anastomosis did not show impairment 10 days after surgery. Nevertheless, it was more impaired in areas identified as having 25 per cent perfusion by FLER compared with 75 per cent perfusion. However, the residual perfusion preserved mitochondrial activity, as demonstrated by the ability to reach comparable Vmax values in both 25 and 75 per cent perfusion areas. No improvement in mitochondrial performance was expected after healing, even in the best perfusion conditions, because only ischaemia was induced, with no reperfusion cycles. An improvement in mitochondrial respiration can be expected if the tissue is preconditioned by means of ischaemia–reperfusion cycles15.
A tracking feature has recently been developed and implemented in the ER software package16. An image-merging modality is also being integrated into the D-Light P, with the aim of obtaining a real-time update of quantitative time-to-peak values. These improvements will decrease the time needed to obtain FLER from the present 4–5 min to around 1 min. Transfer of this technology to the clinical setting is currently being scheduled.
Acknowledgements
The authors are grateful to L. Oudot, C. Burel and G. Temporal for proofreading the manuscript. M.D. is recipient of a research grant from Karl Storz Endoscope. J.M. is the President of both IRCAD and the Institute for Minimally Invasive Image-Guided Surgery, which are partly funded by Karl Storz, Covidien and Siemens Healthcare.
Disclosure: The authors declare no other conflict of interest.
Supporting information
Additional supporting information may be found in the online version of this article:
Video S1 Fluorescence-based enhanced reality for intraoperative assessment of bowel perfusion (wmv file)
Table S1 Exploratory laparotomy findings (Word document)
Table S2 Pathology findings (Word document)
