If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Optimization of anastomotic technique and gastric conduit perfusion with hyperspectral imaging and machine learning in an experimental model for minimally invasive esophagectomy
Department of General, Visceral, and Transplantation Surgery, Heidelberg University Hospital, Heidelberg, GermanyHIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, Germany
Department of General, Visceral, and Transplantation Surgery, Heidelberg University Hospital, Heidelberg, GermanySchool of Medicine, Heidelberg University, Heidelberg, Germany
HIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, GermanyDivision of Computer Assisted Medical Interventions, German Cancer Research Center (DKFZ), Heidelberg, GermanyFaculty of Mathematics and Computer Science, Heidelberg University, Heidelberg, Germany
HIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, GermanyDivision of Computer Assisted Medical Interventions, German Cancer Research Center (DKFZ), Heidelberg, GermanyMedical Faculty, Heidelberg University, Heidelberg, Germany
HIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, GermanyDivision of Computer Assisted Medical Interventions, German Cancer Research Center (DKFZ), Heidelberg, Germany
HIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, GermanyDivision of Computer Assisted Medical Interventions, German Cancer Research Center (DKFZ), Heidelberg, GermanyFaculty of Mathematics and Computer Science, Heidelberg University, Heidelberg, GermanyMedical Faculty, Heidelberg University, Heidelberg, Germany
Corresponding author. Department of General, Visceral, and Transplantation Surgery Heidelberg University Hospital, Im Neuenheimer Feld 420, 69120, Heidelberg, Germany.
Department of General, Visceral, and Transplantation Surgery, Heidelberg University Hospital, Heidelberg, GermanyHIDSS4Health - Helmholtz Information and Data Science School for Health, Heidelberg and Karlsruhe, Germany
Esophagectomy is the mainstay of esophageal cancer treatment, but anastomotic insufficiency related morbidity and mortality remain challenging for patient outcome. Therefore, the objective of this work was to optimize anastomotic technique and gastric conduit perfusion with hyperspectral imaging (HSI) for total minimally invasive esophagectomy (MIE) with linear stapled anastomosis.
Material and methods
A live porcine model (n = 58) for MIE was used with gastric conduit formation and simulation of linear stapled side-to-side esophagogastrostomy. Four main experimental groups differed in stapling length (3 vs. 6 cm) and simulation of anastomotic position on the conduit (cranial vs. caudal). Tissue oxygenation around the anastomotic simulation site was evaluated using HSI and was validated with histopathology.
Results
The tissue oxygenation (ΔStO2) after the anastomotic simulation remained constant only for the short stapler in caudal position (−0.4 ± 4.4%, n.s.) while it was impaired markedly in the other groups (short-cranial: −15.6 ± 11.5%, p = 0.0002; long-cranial: −20.4 ± 7.6%, p = 0.0126; long-caudal: −16.1 ± 9.4%, p < 0.0001). Tissue samples from avascular stomach as measured by HSI showed correspondent eosinophilic pre-necrotic changes in 35.7 ± 9.7% of the surface area.
Conclusion
Tissue oxygenation at the site of anastomotic simulation of the gastric conduit during MIE is influenced by stapling technique. Optimal oxygenation was achieved with a short stapler (3 cm) and sufficient distance of the simulated anastomosis to the cranial end of the gastric conduit. HSI tissue deoxygenation corresponded to histopathologic necrotic tissue changes. The experimental model with HSI and ML allow for systematic optimization of gastric conduit perfusion and anastomotic technique while clinical translation will have to be proven.
Treatment for esophageal cancer is stage-dependent and can include endoscopic resection, esophagectomy and chemotherapy as well as radiation. The mainstay of treatment for any resectable esophageal cancer is esophagectomy [
]. Technical and technological improvements such as minimally invasive surgery and stapled anastomosis have led to reduced morbidity and mortality of esophagectomy over the last decades. There is sufficient evidence nowadays showing that oncological outcomes between open esophagectomy (OE) and minimally invasive technique are at least equivalent [
Robot-assisted minimally invasive thoracolaparoscopic esophagectomy versus open transthoracic esophagectomy for resectable esophageal cancer: a randomized controlled trial.
]. Currently, there is no official gold standard technique for the creation of intrathoracic esophagogastric anastomosis after Ivor-Lewis esophagectomy and the choice of technique depends on the localization of the tumor as well as the surgeon's preference and experience [
]. The perfusion and tissue oxygenation of the gastric conduit and integrity of the anastomosis are critical factors for short- and long-term outcome.
While the end of the gastric conduit can more often experience perfusion-associated problems, perfusion is usually sufficient at the esophageal stump. For the linear-stapled technique in MIE the anastomosis is constructed in a side-to-side fashion. Besides popularity of end-to-side circular stapled anastomosis for OE and robotic-assisted MIE, the linear stapled anastomosis provides a method with good feasibility and potential for standardization in conventional MIE. The linear stapled technique is well established in laparoscopic bariatric surgery and can serve as a mainstay technique for conventional MIE, especially if a robotic system is not available. There is currently no evidence for the exact placement of the linear stapled anastomosis during MIE, although this may influence tissue perfusion and oxygenation and therefore the risk of complications.
Hyperspectral Imaging (HSI) is a novel imaging technique that can estimate tissue oxygenation and microcirculatory perfusion by measuring reflectance intensity separately at different wavelengths [
] (Fig. 1). It then calculates color-coded index images that can be used to measure tissue oxygenation and perfusion and enables microvascular evaluation of organ perfusion [
Hyperspectral imaging as a possible tool for visualization of changes in hemoglobin oxygenation in patients with deficient hemodynamics – proof of concept.
]. It is therefore well suited to evaluate different technical aspects of esophagectomy such as anastomotic stapling techniques regarding tissue perfusion as risk factors for anastomotic leakage.
Fig. 1| Hyperspectral data cube and camera system. a, visualization of a three-dimensional hyperspectral datacube indicating the spectral bands relevant for the calculation of the hyperspectral oxygenation index (StO2). b, TIVITA® Tissue hyperspectral camera system with the kind permission from Diaspective Vision GmbH. c, color-coded picture of StO2 index.
The aim of the present study was to assess different stapling positions and techniques and their effect on tissue oxygenation in a porcine model for MIE. The optimization of gastric conduit perfusion and anastomotic technique in an experimental model for MIE will serve to improve understanding of gastric conduit physiology and to reduce complications for patients requiring esophagectomy.
2. Methods
2.1 Porcine model and surgical procedure
The experiments were approved by the Committee on Animal Experimentation of the regional council Baden-Württemberg in Karlsruhe (G-161/18 and G-262/19). All pigs were managed according to German laws for animal use and care and according to the directives of the European Community Council (2010/63/EU) and ARRIVE guidelines [
In order to obtain a first impression regarding spectral differences between physiological (A = 39; n = 849), avascular stomach (A = 15; n = 117) and stomach with venous congestion (A = 8; n = 80), hyperspectral recordings were taken from physiological stomach, from stomach after complete dissection of connective tissue and blood supply and finally from stomach after artery-preserving dissection (A = 58; n = 1046 in total) (Fig. 2, Fig. 3). A always indicates the number of animals; n always indicates the number of measurements in total.
Fig. 2| Hyperspectral imaging characterization of stomach tissue. TIVITA® color-coded index images for oxygenation (StO2) and relative reflectance intensities across wavelengths. a, images for physiological stomach (A = 39; n = 849). b, images for avascular stomach (A = 15; n = 117). c, images for stomach with venous congestion (A = 8; n = 80). d, StO2 values in comparison. e, reflectance of physiological stomach. f, reflectance of avascular stomach. g, reflectance of stomach with venous congestion. Black arrows indicate the spectral areas that are mainly influenced by the oxygenation status of hemoglobin. A indicates the number of animals; n indicates the number of samples in total. Mean and standard deviation are calculated across individual animals.
Fig. 3| Gastric reflectance in comparison. a, original spectral reflectance in comparison. b, L1-normalized spectral reflectance in comparison. c, PCA of the three baseline groups.
Gastric vascular anatomy was already known from former projects to be highly comparable to human anatomy in the relevant aspects, which also presented itself clearly during the intraoperative preparation. For the purpose of this project's main research question, the specific interest was the anastomotic site at the end of the gastric conduit as this is known to be the critical location regarding perfusion. The anastomotic site at the esophageal stump was not of interest as perfusion is usually sufficient there and hence does not present a regular clinical problem given that mobilization of the esophagus at the height of the anastomosis is kept to the required minimum. Consequently, only the abdominal procedure of Ivor-Lewis esophagectomy was performed as a conventional gastric conduit formation with dissection of the inflow of the left gastroepiploic and left gastric arteries and resection of the lesser curvature [
The procedure was performed in an open fashion via midline laparotomy, but with the technical aspects as used in MIE, i.e. linear stapled anastomosis. Instead of a real linear stapler dissection, we decided to use a simulation in order to have a reversible model and to be able to perform several repeated measurements in one pig for the purpose of number reduction of animals used. The simulation of the linear stapled anastomosis was performed reversibly with a magnet as a linear occlusion of tissue. Only in those cases where tissue integrity could be ensured by clinical inspection and by complete restoration of HSI results to the baseline values, repeated measurements were taken. A standardized gastric conduit was created with stapling devices. For gastric conduit perfusion, the only remaining vessels of supply were the right gastroepiploic artery and vein. During all of the following steps, hyperspectral images of the conduits were recorded.
In order to impair blood supply of the conduit, strong magnets with the same dimensions as typical stapling devices were applied as a linear occlusion model. While one magnet was placed intraluminally, the counterside magnet was placed on the ventral surface of the gastric conduit. Pilot studies were performed to ensure this magnet-induced ischemia was an accurate representation of stapler-induced ischemia and blood flow was completely inhibited (Supplementary Fig. 1). There were magnets in two lengths, namely 3 cm and 6 cm, in order to simulate 3 cm and 6 cm linear stapled anastomoses.
In theory, there are two conceivable ways of perfusion that can contribute to the blood supply of the gastric conduit. These are longitudinal tissue capillary perfusion on the one hand (Fig. 4a) and trans-arterial and subsequent transverse tissue capillary perfusion via the gastroepiploic artery on the other (Fig. 4b). In order to investigate the primary way of blood supply, gastric conduit perfusion was impaired intentionally with transverse and longitudinal magnets. The transverse magnet does not resemble a proper stapling technique and was placed distally in order to maximize the monitoring surface of the proximal gastric conduit. To investigate the four main experimental groups of this work, the linear incision on the ventral side of the gastric conduit usually caused by the linear stapling device in MIE was now simulated with magnets that had the same dimensions as typical stapling devices (Fig. 5).
Fig. 4| Evaluation of possible ways of gastric conduit perfusion with hyperspectral imaging. a, longitudinal tissue capillary perfusion relative to the conduit (A = 10; n = 10) and schematic depiction of relevant gastric anatomy with longitudinal (L) and transverse (T) axis of the later conduit. b, trans-arterial and subsequent transverse tissue capillary perfusion relative to the conduit via the gastroepiploic artery (A = 10; n = 14). Arrows indicate the site of perfusion inhibition through magnets. White boxes indicate regions of interest (ROIs) for measurements. A paired t-test was applied; n.s. is not significant, ∗∗∗ is p ≤ 0.001. Numbers in the box plots were obtained from the ROIs. Different animals are coded in different shapes; circles are different animals. Graphs depict mean and standard deviation.
Fig. 5| Effects of linear stapled esophagogastric anastomotic stapling positions and lengths. I, short (3 cm) cranial stapler simulation (A = 10; n = 14). II: short (3 cm) caudal stapler simulation (A = 9; n = 9). III: long (6 cm) cranial stapler simulation (A = 3; n = 4). IV: long (6 cm) caudal stapler simulation (A = 20; n = 25). a, schematic drawing. b, exemplary visualization of oxygenation and corresponding quantification. White boxes and arrows indicate regions of interest (ROIs) for measurements. For b a paired t-test was performed. c, histograms of all StO2 values without hierarchical structure. d, quantification of the StO2 difference in the ROIs between control and intervention and exemplary color-coded visualization. Black arrows indicate the ROI next to the magnet. For d a one-way ANOVA was performed; only significant differences to group II (short caudal) as reference were highlighted; all other comparisons were not significant; ref indicates reference group for statistical testing. ∗ is p ≤ 0.05, ∗∗ is p ≤ 0.01, ∗∗∗ is p ≤ 0.001, ∗∗∗∗ is p ≤ 0.0001, n.s. is not significant. Numbers in the box plots were obtained from the ROIs. e, direct comparison of StO2 value visualization for the 3 cm and 6 cm group. f, PCA of the three baseline groups and experimental group IV as the largest uniform experimental group that also provided the pathohistological samples. g, spectral comparison of the three baseline groups and experimental group IV.
The four main experimental groups that represent possible stapling methods for the linear stapled anastomosis were (group I) short stapler simulation (3 cm length) with cranial stapling position, (group II) short stapler simulation with caudal stapling position, (group III) long stapler simulation (6 cm length) with cranial stapling position, and (group IV) long stapler simulation with caudal stapling position (Fig. 5a). While the cranial stapling position describes the closest possible position of the intraluminal magnet to the cranial staple line of the gastric conduit with no well perfused tissue remaining at the end of the conduit, the caudal stapling position describes the magnet position 3 cm below, so that there is 3 cm of well perfused gastric conduit tissue still above. HSI recordings were done for the physiological gastric conduit and then with the magnets in place to simulate the anastomosis after 2 min.
2.2 Hyperspectral imaging and statistical analysis
HSI data was acquired with the TIVITA® Tissue system (Diaspective Vision GmbH) (Fig. 1). It provides a spectral resolution of 5 nm in the range from 500 nm to 995 nm. Six integrated halogen lamps provide a standardized illumination. Numerical values stored in the datacube represent reflectance values at every single pixel for every wavelength in arbitrary units. The computed tissue parameter index images include the hyperspectral oxygenation index (StO2) and underlying formulas can be reviewed in cited literature [
]. For visualization purposes, the control StO2 images were subtracted from the intervention StO2 images, resulting in ΔStO2 images.
Mean and standard deviation (SD) reflectance spectra of baseline stomach recordings across animals were obtained by calculating the median spectrum over every pixel within the annotated region of interest (ROI) of one image and then the mean spectrum over every image within one pig. For the comparison of gastric conduit recordings before (control) and after magnet application for anastomosis simulation (intervention) both were registered using the Optical flow algorithm based on the annotation of the contour of each conduit [
] (Supplementary Fig. 2), a ROI of identical size (16 × 50 pixels) was manually placed for each recording in the intervention image and the ROI in the control image was set automatically according to the registration. For any figures reporting StO2, the mean ± SD over every pixel within the ROI of one image was calculated and then the mean over every image of one experimental group. For the analysis of the four main experimental groups, a ROI for each recording in the intervention group was defined on the left side of the magnet. Baseline measurements were taken from respective registered control recordings without magnet.
All displayed StO2 values were obtained as relative values from the official formula used by the HSI camera as cited in the literature [
]. Different animals were coded in different shapes; in analyses of hyperspectral data circles are measurements from different animals. A p-value ≤0.05 was considered statistically significant. In case of parametric data, paired and unpaired t-test was used. In case of non-parametric data, Mann-Whitney test was used for unpaired and Wilcoxon matched-pairs signed rank test for paired data. For comparisons of multiple groups, one-way ANOVA was used in case of parametric normal distribution while Kruskal-Wallis was used in case of non-parametric distribution. p-values were adjusted for multiple testing. Scale bars always equal 5 cm if not stated otherwise. Machine Learning (ML) in the form of principal component analysis (PCA) was used to visualize spectral differentiability between the three baseline groups as well as identify affiliation of the magnet-induced gastric ischemia. Graph depicts mean and standard deviation.
2.3 Vascular corrosion casting and scanning electron microscopy
Vascular corrosion casting of the stomach was obtained using Biodur E20® (Biodur Products, Heidelberg, Germany) in order to investigate possible rheologic-mechanistic explanations for HSI findings. After euthanasia with i.v. potassium chloride, the abdominal aorta was cannulated, the distal thoracic aorta and iliac arteries were clamped and the inferior caval vein was incised. Perfusion was initiated by pressured infusion of 1000 ml of Sterofundin® (B. Braun®) through previously mentioned cannulation including 50,000 I.U. heparin, followed by 400 ml Sterofundin®. Biodur E20® for injection was mixed at a ratio of 100:45 (v/v) Biodur E20® Plus and catalyst E20 and injected by manual pressure. The infusion was stopped after venous return of the casting agent was observed in the inferior caval vein and the material initially hardened for several minutes without further manipulation. The gastric specimen was explanted and incubated for 12 h in a 40 °C water bath for further hardening. Gastric tissue was removed with 15% (w/v) potassium hydroxide (RT; 3 days) and the resulting vascular corrosion cast was subsequently rinsed in water. Scanning electron microscopy (SEM) samples were prepared from corpus areas of stomach corrosion cast specimens. 20 mm × 10 mm samples of the outer layer were 10 nm gold/platinum (80:20) sputtered (Leica EM ACE 600, Leica Microsystems GmbH, Wetzlar, Germany) and analyzed by scanning electron microscopy (Zeiss Leo Gemini 1530, Carl Zeiss AG, Oberkochen, Germany). SEM images were taken at different magnifications with an accelerating voltage of 2.0 kV.
2.4 Pathohistological analysis
In order to evaluate whether the differences in HSI oxygenation have any biological relevance and reflect histopathological changes, tissue in the main experimental group IV was sampled from physiological stomach (R1) and from 3 distinctive regions after 6 h of magnet-induced ischemia (R2-4). R2 serves as baseline of physiological tissue 6 h into the surgery. R4 is defined by lowest HSI oxygenation alongside the magnet. R3 is the intermediate region. All tissue samples were formalin-fixated and hematoxylin-eosin stained. Slides were scanned using widefield microscopy (Zeiss) and the amount of mucosal eosinophilic pre-necrotic areas was objectified with the specialized software DeePathology™ STUDIO (Deepathology Ltd., Israel).
3. Results
3.1 Physiological and avascular stomach
Differences between physiological stomach (72.4 ± 9.2%), avascular stomach (34.6 ± 6.1%) and stomach with venous congestion (65.5 ± 8.2%) could be clearly detected in the HSI StO2 images (Fig. 2).
Distinct changes in the reflectance spectrum were seen between 600 and 650 nm and above 700 nm. Physiological and avascular stomach could be well differentiated by PCA with an explained variance of 89.25%. Stomach with venous congestions has greater similarity to physiological stomach, but still forms a specific cluster (Fig. 3).
3.2 Capillary and gastroepiploic perfusion
Inhibiting longitudinal tissue capillary perfusion by using a transverse magnet position (Fig. 4a) did not yield any significant difference between the StO2 values of control (72.7 ± 6.1%) and intervention (74.8 ± 4.6%). However, when limiting transarterial and subsequent transverse tissue capillary perfusion via the gastroepiploic artery with a longitudinal magnet (Fig. 4b), StO2 levels significantly dropped from 66.2 ± 6.9% to 50.6 ± 7.9% (p = 0.0002). Furthermore, it could be seen that gastric conduit perfusion directly depended on the height of gastroepiploic occlusion (Supplementary Fig. 3).
3.3 Different stapler simulation sizes and positions influence tissue oxygenation of linear stapled anastomostic simulation during esophagectomy
Due to results of aforementioned experiments regarding capillary and gastroepiploic perfusion, the region with the lowest StO2 values on the gastric conduit during intervention can be expected on the left side of the magnet downstream to the perfusion coming transversely from the gastroepiploic artery. The four main experimental groups representing possible linear stapling simulation methods for constructing the anastomosis during MIE were evaluated for changes in tissue oxygenation (Fig. 5). Oxygenation levels were significantly reduced in group I (short-cranial) from 66.2 ± 6.9% to 50.6 ± 7.9% (ΔStO2 = −15.6 ± 11.5%, p = 0.0002), in group III (long-cranial) from 59.6 ± 4.5% to 39.3 ± 9.0% (ΔStO2 = −20.4 ± 7.6%, p = 0.0126) and in group IV (long-caudal) from 60.3 ± 13.0% to 44.3 ± 7.7% (ΔStO2 = −16.1 ± 9.4%, p < 0.0001) (Fig. 5b.I, III and IV). Only in group II (short-caudal) changes from 70.5 ± 6.4% to 70.1 ± 9.0% were not significant (ΔStO2 = −0.4 ± 4.4%, n.s.) (Fig. 5b.II). ΔStO2 images (Fig. 5d) were all significantly different from group II, which served as a reference with values of −0.4 ± 4.4%. ΔStO2 values were −15.6 ± 11.5% (p = 0.0021) for group I, −20.4 ± 7.6% (p = 0.0043) for group III and −16.1 ± 9.4% (p = 0.0004) for group IV. All other comparisons were not significantly different. This allows for the conclusion that only for a short stapler simulation with caudal position, tissue oxygenation does not significantly drop after tissue dissection. Fig. 5e shows that this is not a continuous process as one might expect, but rather a discrete change. For the 3 cm simulation loss of perfusion can be completely compensated through random pattern perfusion, but for the 6 cm simulation there are sites that are unambiguously perfused insufficiently. Moreover, ML - using PCA with an explained variance of 89.17% - could show that in general the critical region on the gastric conduit after stapler simulation did not exclusively have an arterial perfusion problem, but also a significant venous congestion component (Fig. 5f). Additional investigation of the influence of conduit width can be found in Supplementary Fig. 4.
3.4 Anatomical correlation to hyperspectral findings
Evaluation of hyperspectral findings regarding tissue oxygenation suggests that the blood supply mainly comes from the gastroepiploic artery and travels transversely through the gastric conduit as opposed to longitudinally from former gastric antrum to gastric cardia. Therefore, the question arose, whether an anatomical correlation providing a pathomechanistic explanation to these results could be found. Corrosion casting of 2 porcine stomachs revealed macroscopically highly comparable relevant vascularity between human and porcine stomach according to surgical experts. After scanning electron microscopy of cast gastric capillaries, it could be quantified that a significantly greater share of all vascular pixels (Fig. 6c) had a greater length of interrupted vascular neighbor pixels in the transversal orientation (67.4% ± 10.5%) than in the longitudinal orientation (32.6% ± 10.5%) (p < 0.0001) (Fig. 6). This not only supports the hypothesis of predominantly transgastroepiploic blood supply of the conduit as it was done in previous anatomical studies in patients [
], but it moreover delivers pathomechanistic explanations to the observed phenomenon of capillary orientation.
Fig. 6| Corrosion casting of porcine stomach and scanning electron microscopy (SEM). a, cast porcine stomach. b, magnification with assumed blood flow via gastroepiploic artery as indicated by red arrows. c, manual annotation of vessels from the median gastric corpus wall. d, quantification of vessel orientation separated in primarily transverse (from lateral to medial gastric edge) and longitudinal (from fundal to antral gastric edge) orientation depending on the longest distance of uninterrupted pixels in the annotation (A = 2; n = 9). e, SEM sample from median gastric corpus. f, SEM sample from lateral gastric corpus with visible gastroepiploic artery. Unpaired t-test was performed; ∗∗∗∗ is p ≤ 0.0001.
3.5 Pathohistological correlation to hyperspectral findings
The area of pre-necrotic changes in the samples of 4 porcine stomachs were quantified with 5.5 ± 3.0% for R1, 8.3 ± 3.2% for R2, 18.1 ± 10.0% for R3 and 35.7 ± 9.7% for R4 (Fig. 7). All differences were significant as indicated (p < 0.0001).
Fig. 7| Pathohistological samples in correlation to hyperspectral findings. a-d, Pathohistological HE samples showing whole slides and representative mucosal areas for R1: physiological (A = 4; n = 54), R2: well perfused for 6h (A = 4; n = 41), R3: critically perfused for 6h (A = 4; n = 42) and R4: not perfused for 6h (A = 4; n = 36). e, hyperspectral recording of gastric conduit with impaired perfusion due to magnet-induced ischemia. f, DeePathology™ STUDIO with visualization of physiological areas in magenta and eosinophilic pre-necrotic areas in green. g, quantification of eosinophilic pre-necrotic areas in percentage of surface area. Ordinary one-way ANOVA with multiple comparisons was used. n.s. is not significant, ∗∗∗∗ is p ≤ 0.0001. Upper scale bar indicates 1 mm, lower scale bar indicates 500 μm.
Anastomotic integrity is the Achilles’ heel of esophagectomy and other gastrointestinal procedures. It is multifactorial with the most important influencing factors being sufficient tissue perfusion, lack of tension, asepsis and quality of surgical sutures and stitches [
]. The present experimental study investigated several technical aspects of linear stapled esophagogastric anastomosis used in Ivor Lewis MIE focusing on anastomotic perfusion. Linear stapled side-to-side esophagogastric anastomosis is a viable alternative to open or robotic circular stapled technique and especially valuable if advantages of MIE are desired despite the lack of a robotic system [
Minimally Invasive versus open AbdominoThoracic Esophagectomy for esophageal carcinoma (MIVATE) - study protocol for a randomized controlled trial DRKS00016773.
]. In bariatric surgery the linear stapled side-to-side anastomosis has proven to be a reproducible and standardizable technique with good clinical performance causing it to be the focus of this investigation with regard to MIE [
Comparison between circular- and linear-stapled gastrojejunostomy in laparoscopic Roux-en-Y gastric bypass--a cohort from the Scandinavian Obesity Registry.
]. The present study established a translational method and blueprint for the systematic evaluation of different technical aspects of the gastric conduit and anastomosis for esophagectomy in a porcine model with magnet-based anastomotic simulation. HSI was used as a tool for objective tissue perfusion evaluation intended for optimization of surgical technique.
Spectral reflectance was clearly different between physiological and non-perfused gastric tissue with a double-peak between 550 and 590 nm and higher values at 680 nm for physiological tissue compared to non-perfused gastric tissue as seen in Fig. 3. These findings are in line with the descriptions of optical properties of oxygenated and deoxygenated hemoglobin in current literature [
]. This initial analysis serves as a reference point for the subsequent experiments as spectral reflectance of the critically perfused areas on the gastric conduit next to the stapling simulation site is expected to be similar compared to the spectral reflectance of non-perfused gastric tissue.
As a first step, basic principles of gastric conduit perfusion dynamics were explored by impairing the two possible ways of blood supply i.e. longitudinal capillary perfusion in the gastric conduit tissue along the greater curvature or alternatively perfusion along the gastroepiploic artery and subsequently capillaries transversely to the conduit axis as illustrated in Fig. 4. It could be shown that there is no sign of relevant longitudinal capillary perfusion, but proof of perfusion along the gastroepiploic artery as seen in Fig. 4, Supplementary Figs. 1 and 3. This observation matches results from human anatomical studies, which showed that larger capillaries are distributed perpendicular to both gastric curvature sides and only small capillary branches parallel to both curvature sides originate from these to form a vascular network throughout the stomach wall [
]. The dominant direction of flow is therefore not a random pattern along the longitudinal axis of the conduit. Thus, perfusion mainly occurs through the right gastroepiploic artery and subsequent capillaries along the transverse axis of the conduit. Especially the distal end of the conduit also relies on a sufficient collateral connection between right and left gastroepiploic artery. Impairment in this collateral anatomy might cause insufficient perfusion of the conduit's end with the consequences of necessary resection and shortening possibly leading to anastomotic tension. A human study found the right gastroepiploic artery to be more dominant in that the mean cross-sectional area was found to be 3.31 ± 1.71 mm2 compared to 1.33 ± 1.01 mm2 for the left gastroepiploic artery and 0.51 ± 0.28 mm2 for the collateral connection between both [
]. Similar observations were made during the present porcine study in the sense that the right gastroepiploic artery was the dominant vessel supplying the gastric conduit. Furthermore, different human anatomical studies report different rates of actual collateral anastomotic connections between right and left gastroepiploic artery reaching from as high as 95% with an arterial arch (70%) or mesh-like anastomosis (25%) [
]. The knowledge of these anatomical key points is crucial for gastric conduit geometry and formation. It can therefore be concluded that in the porcine model for esophagectomy with reconstruction via a gastric conduit, the relevant amount of perfusion is almost exclusively due to the right gastroepiploic artery and its transverse capillary branches. The possible length of the gastric conduit therefore also depends on size and collateral connection between both gastroepiploic arteries.
An in-depth analysis of different anastomotic linear stapler simulation techniques showed that the most promising stapling technique in order to preserve the optimal tissue perfusion measured with HSI is a short stapler (3 cm as compared to 6 cm) with sufficient distance (3 cm) of the cranial end of the anastomosis to the stapled end of the conduit as depicted in Fig. 5. The exact distances and sizes of staple lines are not always reported in clinical practice and no publications could be found addressing the proximity of the anastomosis towards the cranial end of the conduit. However, there were two comparable clinical trials investigating MIE Ivor Lewis esophagectomy of which one used a 6 cm linear stapler [
]. Within the 124 patients with a 6 cm anastomosis 9 anastomotic leaks occurred (7.3%), while within the 104 patients with a 3 cm anastomosis only 4 patients (3.8%) experienced anastomotic leakage. Overall postoperative complications were similar with 51.6% compared to 50.0%. Although one could expect higher stricture rates for a smaller stapler, the 3 cm anastomosis only had 1 stricture (<1.0%) compared to the 6 cm anastomosis with 6 patients (5.1%) with anastomotic stricture requiring endoscopic intervention. This is in line with the current study indicating that hypoperfusion of the perianastomotic tissue is more prevalent with longer staplers used for anastomosis, which can lead to anastomotic insufficiency on the short term, but also to scarring and strictures on the long term. It therefore seems that patients with a shorter linear stapled anastomosis benefit from improved anastomotic perfusion with potentially lower risk of anastomotic leakage while not bearing additional risks for stricture.
Since HSI data is multidimensional per definition and exceeds the comprehension capacity of the human brain, making use of artificial intelligence analyses and ML seems particularly useful. And indeed, as feasibility and usability could be shown in previous publications [
], also in this specific use case ML was able to extract valuable insights regarding the role of venous congestion for this simulated linear stapler anastomotic technique.
In order to show biological relevance and potential for clinical translation of the HSI-based StO2 measurements, the effect on tissue necrosis was investigated by histopathological examination in the present study. Differences between gastric conduit groups in histologically-observed tissue reactions to induced tissue ischemia clearly corresponded to the observed different hyperspectral StO2 values. It could be shown that lower StO2 values in HSI corresponded to higher percentages of pre-necrotic stains in the gastric mucosa of histopathological samples, hereby indicating significant tissue damage through insufficient perfusion (Fig. 7). Physiologically high StO2 values corresponded to physiologically low levels of pre-necrotic tissue changes. A typical drop in StO2 from physiological HSI values between 60% and 70% as described in other studies [
Quantitative serosal and mucosal optical imaging perfusion assessment in gastric conduits for esophageal surgery: an experimental study in enhanced reality.
] to below 45% during magnet-induced ischemia resulted in a pre-necrotic demarcation of gastric tissue of 35.7 ± 9.7% of histological surface area. This tendency also applied incrementally for the stages in between, so that with lower StO2 values between 45% and 60%, there were correspondingly greater surface areas between 10% and 35% with pre-necrotic changes. Histopathological analysis as seen in Fig. 7 therefore confirmed the relevance of HSI StO2 changes for assessment of real tissue condition.
There are several limitations of this study. First, this study only investigated simulations of linear stapled anastomoses as opposed to circular stapled anastomoses for OE. What motivated to place the focus on the investigation of linear stapled anastomoses was the applicability to MIE as well as the excellent experience obtained specifically for this type of anastomoses in bariatric surgery. The study also exclusively investigated the anastomotic simulations so that no actual linear staplers for anastomoses were used or anastomoses were constructed as this would have been counterproductive in obstructing the view on the tissue during measurements.
Secondly, the HSI system used in this study or HSI in general is not capable of recording in real-time due to the large amount of data across 100 wavelengths per measurement. However, this does not compromise the validity of the data, but only limits the applicability during regular surgery to a certain degree, since it currently takes several seconds to perform one measurement intraoperatively as a static image. Yet, there are several human trials that indicate applicability during regular surgery, provided there is compliance from the surgeon [
], which only records a smaller number of wavelengths, but in a fraction of the time and although information content is much lower, it might be more suitable for specific indications with in this respect optimized wavelength constellation.
As another limitation, it also has to be considered that all of the experiments were performed in a porcine model and although the porcine anatomy is very similar to the human anatomy, clinical translation of these results has to be proven. There have been other publications investigating gastric conduit stapling techniques in patients [
]. And while pigs do have a very prominent left gastroepiploic vessel supplying the vascular arcade along the greater curvature that can sometimes be sparse in humans, for gastric conduit formation this left gastroepiploic will be ligated so that the vascular arcade again relies entirely on the right gastroepiploic artery and therefore ensures conformity of the model and sufficient comparability to human patients.
Furthermore, for patients receiving esophagectomy there are additional factors that influence anastomotic integrity. While ideal stapling position and gastric conduit formation as described in this work may represent one decisive factor for anastomotic healing, other factors also come into play such as a tension-free anastomosis, fluid volume management, vasopressors, blood pressure, respiratory function and resulting oxygen supply, comorbidities and nutrition amongst others. It is therefore essential for the successful treatment of patients to take all the relevant factors into consideration in order to obtain the best possible outcome. Explicitly for the correlation of HSI StO2 values and histopathological tissue changes, it is important to mention that this type of validation will have to be repeated for different hyperspectral index picture values, organs and even species.
These limitations are essential, but when accounted for, the present animal model for MIE can be used as a blueprint for the systematic investigation of gastric conduit perfusion dynamics for multiple research questions yet to come.
5. Conclusion
A porcine model for optimization of gastric conduit formation with analysis of tissue oxygenation and microcirculation using the novel intraoperative imaging technique of HSI and use of ML was developed. Different aspects of anastomotic linear stapling technique for minimally invasive Ivor Lewis esophagectomy were evaluated. A shorter linear stapler simulation (3 cm compared to 6 cm) with at least 3 cm distance of the anastomosis to the cranial end of the gastric conduit provided optimal results in terms of tissue oxygenation and perfusion and is thus the most promising technique in terms of anastomotic healing in a porcine model. This animal model with HSI and ML can be used as a blueprint for further systematic investigation of gastric conduit perfusion and optimization of anastomotic technique. Clinical trials are essential to investigate translation to optimized clinical outcome with potential reduction of anastomotic insufficiency and conduit necrosis.
Funding
Willy Robert Pitzer Foundation (grant number: not applicable)
Heidelberg Foundation of Surgery (grant number: not applicable)
Disclosures
Alexander Studier-Fischer, Berkin Özdemir, Jan Odenthal, Lucas-Raphael Müller, Samuel Knödler, Karl-Friedrich Kowalewski, Isabella Camplisson, Michael Martin Allers, Maximilian Dietrich, Karsten Schmidt, Gabriel Alexander Salg, Hannes Götz Kenngott, Adrian Theophil Billeter, Ines Gockel, Chen Sagiv, Ofir Etz Hadar, Jacob Gildenblat, Leonardo Ayala, Silvia Seidlitz, Lena Maier-Hein and Beat Peter Müller-Stich have no conflicts of interest or financial ties to disclose. Felix Nickel reports support for courses and travel from Johnson and Johnson, Medtronic, Intuitive Surgical, Cambridge Medical Robotics and KARL STORZ as well as consultancy fees from KARL STORZ.
Research funding
There was financial support from the Willy Robert Pitzer Foundation and the Heidelberg Foundation of Surgery for this project. Medtronic provided the stapling devices. Conflict of interest: Authors state no conflict of interest. Published data will be made available upon reasonable request to the corresponding author.
The content of this manuscript has been posted on the bioRxiv preprint server on the 4th of October 2021 under the digital object identifier https://doi.org/10.1101/2021.10.03.462901 [
Optimization of anastomotic technique and gastric conduit perfusion with hyperspectral imaging in an experimental model for minimally invasive esophagectomy.
]. The copyright holder for this preprint is the author who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.
CRediT authorship contribution statement
F. Nickel: Conceptualization, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition. A. Studier-Fischer: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. B. Özdemir: Software, Investigation, Data curation. J. Odenthal: Methodology, Software, Data curation. L.R. Müller: Methodology, Validation, Formal analysis, Data curation, Writing – original draft, Visualization. S. Knoedler: Investigation. K.F. Kowalewski: Investigation. I. Camplisson: Methodology, Software. M.M. Allers: Investigation. M. Dietrich: Investigation. K. Schmidt: Investigation, Writing – review & editing. G.A. Salg: Investigation, Writing – review & editing. H.G. Kenngott: Investigation, Writing – review & editing. A.T. Billeter: Investigation, Writing – review & editing. I. Gockel: Conceptualization, Writing – review & editing. C. Sagiv: Conceptualization, Writing – review & editing. O.E. Hadar: Conceptualization, Writing – review & editing. J. Gildenblat: Conceptualization, Writing – review & editing. L. Ayala: Methodology, Software, Formal analysis, Data curation, Visualization. S. Seidlitz: Methodology, Software, Validation, Formal analysis, Data curation, Writing – original draft, Visualization. L. Maier-Hein: Conceptualization, Methodology, Software, Resources, Writing – review & editing, Supervision, Project administration. B.P. Müller-Stich: Conceptualization, Resources, Writing – review & editing, Supervision, Project administration.
Declaration of competing interest
The authors declare that there is no conflict of interest for the submission of the manuscript “Optimization of anastomotic technique and gastric conduit perfusion with hyperspectral imaging and machine learning in an experimental model for minimally invasive esophagectomy”.
Acknowledgements
The authors gratefully acknowledge the data storage service [email protected] supported by the Ministry of Science, Research and the Arts Baden-Württemberg (MWK) and the German Research Foundation (DFG) through grant INST 35/1314-1 FUGG and INST 35/1503-1 FUGG. Furthermore, the authors gratefully acknowledge the support from the Willy Robert Pitzer Foundation and the Heidelberg Foundation of Surgery as well as the support from Medtronic by providing the stapling devices. For the publication fee we acknowledge financial support by Deutsche Forschungsgemeinschaft within the funding programme Open Access Publikationskosten“ as well as by Heidelberg University.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
Robot-assisted minimally invasive thoracolaparoscopic esophagectomy versus open transthoracic esophagectomy for resectable esophageal cancer: a randomized controlled trial.
Hyperspectral imaging as a possible tool for visualization of changes in hemoglobin oxygenation in patients with deficient hemodynamics – proof of concept.
Minimally Invasive versus open AbdominoThoracic Esophagectomy for esophageal carcinoma (MIVATE) - study protocol for a randomized controlled trial DRKS00016773.
Comparison between circular- and linear-stapled gastrojejunostomy in laparoscopic Roux-en-Y gastric bypass--a cohort from the Scandinavian Obesity Registry.
Quantitative serosal and mucosal optical imaging perfusion assessment in gastric conduits for esophageal surgery: an experimental study in enhanced reality.
Optimization of anastomotic technique and gastric conduit perfusion with hyperspectral imaging in an experimental model for minimally invasive esophagectomy.
00444-4&id=gr1.jpg){kind=link}
00444-4&id=gr2.jpg){kind=link}
00444-4&id=gr3.jpg){kind=link}
00444-4&id=gr4.jpg){kind=link}
00444-4&id=gr5.jpg){kind=link}
00444-4&id=gr6.jpg){kind=link}
00444-4&id=gr7.jpg){kind=link}