Esophagectomy is a major and morbid surgical procedure, which plays a central role in the multidisciplinary treatment of patients affected by esophageal neoplasia [
1]. Postoperative complications negatively affect long-term oncological survival [
2]. Anastomotic leakage (AL) is a particularly severe complication, which has a higher incidence following esophagectomy when compared to other digestive surgical procedures [
3]. Although the etiology of anastomotic leaks (AL) is multifactorial, a well-identified risk factor is an insufficient perfusion at the anastomotic site. In fact, the resected esophagus is frequently replaced by a gastric conduit (GC), which is obtained from the tubulation of the stomach, which involves a sudden disruption of most of its blood supply, resulting in an unsatisfactory anastomotic perfusion.
The appreciation of clinical parameters, such as a serosal color or the presence of pulse, is inadequate to rule out a marginally perfused anastomosis [
4]. Several technologies have been explored to intraoperatively assess gastrointestinal tract perfusion. However, the majority of them either lacks reproducibility, disrupts the surgical workflow, or provides operator-dependent results [
5]. Fluorescence angiography (FA), a user-friendly technology which requires a near-infrared camera and the intravenous injection of an exogenous fluorophore, has shown promising results [
6‐
8]. However, the absence of a validated quantification method of the fluorescence signal is one of the main drawbacks to be solved before understanding the real impact of this technique on anastomotic complications. Additionally, the need for injecting a contrast agent represents a regulatory burden, which in practice hinders the diffusion of this technique beyond research protocols and its introduction as a standard of care.
Hyperspectral imaging (HSI) is a contrast-free technology, which can provide a real-time snapshot of the chemical characteristics of a tissue. In particular, this imaging technique allows for tissue oxygen saturation (StO
2) quantification and its utility during gastrointestinal procedures has been previously assessed [
9‐
11]. The main shortcoming of HSI is the lack of video rate, which de facto greatly limits its usability as a surgical navigation tool. Our group previously described and validated the HYPER (HYperspectal-based Enhanced Reality) tool. HYPER results from the superimposition of the HSI-generated perfusion quantification pseudo-color map onto a real-time video of the surgical scene, thereby allowing for a precise spatial localization of the spectral information during surgical procedures. HYPER strongly correlated with robust ischemia parameters in an experimental small bowel and liver partial ischemia models [
12,
13]. However, since HSI is an optical imaging modality, it exclusively allows for a superficial tissue analysis. As a result, serosal perfusion only can be quantified. It has been previously shown that the ischemia behavior within the gastrointestinal tract varies when comparing the serosal with the mucosal side [
14]. Confocal laser endomicroscopy (CLE) is a high-resolution microscopic optical imaging modality, which allows for a detailed morphological assessment of the gastrointestinal mucosa. CLE potentially provides a precise snapshot of the tissue microcirculation, simultaneously giving a morphological in vivo histopathological appraisal and computing the functional capillary density area (FCD-A) index or the speed of red blood cells [
15‐
17].
The aim of the current study was twofold: first, to demonstrate the accuracy of HYPER in a porcine gastric conduit model; second, to compare HSI serosal StO2 with CLE-based mucosal functional capillary density, using the ischemia surrogate, local capillary lactate (LCL), as “ground truth.”
Discussion
In the current study, it was possible to use HYPER in order to precisely assess gastric conduit perfusion immediately after its creation. The enhanced reality tool allowed for an accurate spatial localization of the StO
2 information, and this subsequently allowed to scan the gastric mucosa with CLE and sample the LCL in exact correspondence of each ROI. Additionally, we observed a lower tissue perfusion, with both optical imaging techniques, at the gastric fundus (ROI-F), when compared to the gastric body (ROI-C) and to the pylorus (ROI-P). The relative ischemia present at the fundic region was also confirmed by the significant increase in LCL values within this ROI, in comparison with the remaining ROIs. These findings demonstrate that the future anastomotic region on the gastric conduit suffers from a relative ischemia and this is consistent with previous animal [
24] and clinical [
25,
26] studies in which perfusion was assessed during different phases of esophageal resection. Although we did not assess baseline gastric perfusion routinely in our study, we can assume that a perfusion drop from baseline levels occurred after gastric tube completion, as observed by others [
24,
25]. Accordingly, in the example shown in Fig.
1, StO
2 dropped consistently at ROI-F after slender tube creation of the stomach, whereas it changed only marginally in the other ROIs.
The marginal perfusion of esophagogastric anastomoses, resulting from a sudden disruption of the majority of the gastric fundus blood supply, is an open critical issue in digestive surgery. Consequently, in the past, several authors have used different tools to assess gastric blood flow intraoperatively, in an attempt to identify poorly perfused areas and possibly prevent anastomotic complications. Budlau et al. used a special endoscopic probe which could simultaneously act as a Doppler flow meter and spectrometer, hence allowing the concurrent quantification of mucosal blood flow and oxygen saturation [
27]. In a feasibility study, the authors tested the usability of this probe in healthy individuals undergoing gastroscopy for several reasons. In a second study, the same authors successfully quantified mucosal oxygen saturation intraoperatively during esophagectomy by means of a similar spectrometric probe [
26]. Recently, Irino et al. used a pulse oximetry device attached to an ear probe to assess the serosal oxygen saturation of the stomach during different steps of esophagectomy [
25]. However, all of those probe-based optical techniques provide a pin-point perfusion quantification with none or minimal spatial information, and this greatly limits their routine intraoperative use. In this view, technologies with a better spatial resolution, such as laser speckle [
28], fluorescence angiography (FA) [
6], or thermography [
29] have been explored. However, in spite of interesting results, the laser speckle’s perfusion measurement is highly variable in relation to distance and angulation from the target, and the accuracy of this technology has not yet been validated in a controlled experimental setting, using reliable ischemia biomarkers. On the other hand, FA is an easy-to-use technology, which uses an exogenous fluorophore and special near-infrared (NIR) cameras, which are becoming increasingly available in the OR, both as open and laparoscopic cameras. This technique is effective in assessing GC blood flow [
6,
8,
30]. However, the evaluation is mostly based on a qualitative assessment, limiting FA reproducibility. Several FA quantification methods have been proposed [
31,
32], showing promising results. However, those techniques have not been validated using robust biological ischemia markers, and as a result, their actual correlation with tissue perfusion remains unclear. Recently, a previously experimentally validated FA quantification method has been used to measure GC perfusion [
33]. This quantitative FA is based on a very similar algorithm to the one previously developed and validated by our group [
16,
21,
22], which computes the speed required by the injected fluorophore to reach its maximal intensity peak pixel-by-pixel. The results of this method are promising. However, this metric is not yet universally accepted and larger trials are necessary to understand its real clinical usefulness. Nishikawa et al. [
29] have shown interesting results using thermal imaging during esophagectomies. In particular, the authors proposed a score based on the temperature and on the length of the gastric conduit and its supplying vessels. In their large series, they observed that patients with a lower score were more likely to develop anastomotic leakage. However, in our opinion, such a score, despite remarkable results, might be too complex to routinely reproduce outside of a research protocol.
Previously, our group used hyperspectral imaging to assess StO
2 intraoperatively during esophagectomy [
11]. In a pilot study, the feasibility of the technique was assessed, without using HYPER and subsequently only having a limited intraoperative colocalization of the perfusion information. In the current experimental study, we used HYPER, in order to localize each ROI intraoperatively, with a high degree of accuracy, and subsequently to perform a mucosal scan with CLE and LCL sampling on similar exact spots. HYPER allowed us to compare both optical techniques with one another and with local capillary lactate levels. Interestingly, our findings highlighted a good correlation between mucosal and serosal perfusion assessed using HSI and CLE, respectively. The metrics obtained by both imaging techniques correlated negatively with local capillary lactate values, confirming the biological validity of both methods. CLE correlated strongly and better than HSI to LCL levels. This finding was also confirmed by LCL prediction models, in which the FCD-A index-based prediction was significantly better than the StO
2-based model. On the one hand, this could be explained by a better accuracy of CLE than HSI in determining blood perfusion. On the other hand, it has been demonstrated that ischemia is more extended at the mucosal than at the serosal side [
14], and it is also possible that, during our experiment, mucosal ischemia was more severe and subsequently correlated better to LCL values than to the serosal value. However, this is merely a hypothesis, which cannot be demonstrated by our experimental setup. Interestingly, as previously observed [
12,
19], by excluding the clearly ischemic ROIs (the ones presenting LCL values > 6 mmol/L), the precision of both LCL prediction models greatly increased. The computed prediction models excluding the highly ischemic ROIs showed no statistical difference between them. The poor accuracy of HSI to detect very ischemic regions does not affect its clinical utility, since those areas are generally well-recognized visually, and the regions difficult to assess are the ones marginally perfused, which were recognized by means of HSI with good precision.
CLE is not meant to assess bowel ischemia, since it has been conceived to perform a high-resolution in vivo histopathological assessment. However, its efficacy in identifying porcine bowel ischemia, based on micro-imaging morphological changes [
16] or on the FCD-A index [
17], had been previously demonstrated. In our study, CLE also proved to be a very precise tool to assess gastrointestinal perfusion. However, despite these promising results, the microscopic field of view greatly limits the usability of this device as an intraoperative perfusion assessment tool, as for all probe-based devices.
HSI is a relatively easy-to-use technique, which requires an additional camera (hyperspectral imager) providing the blood flow information as an immediate output. As previously described, this technology does not impair the surgical workflow during esophagectomy [
11], since it is contactless, non-invasive and the acquisition time is less than 10 s. Differently, the intraoperative use of CLE in order to assess the mucosal perfusion of the gastric graft could be more troublesome. Firstly, as mentioned above, this technology has a microscopic spatial resolution, providing pin-point information. Secondly, it does not provide immediate perfusion results, requiring a videos post-processing. Thirdly, its utilization in the OR would either imply the intraoperative use of flexible endoscopy, with the potential hazard of manipulation of a recent esophago-gastrostomy or would require performing a gastrotomy on the GC, as in our experimental setup. Both CLE endoluminal routes are time consuming and would theoretically affect the surgical workflow; therefore, we believe that despite its greater precision and its experimental interest, CLE might not be suitable for routine intraoperative use.
The main drawbacks of our study lie in the very limited sample size, in the animal model and in the acute nature of the experiments. However, our study has the merit to demonstrate for the first time the positive strong correlation existing between hyperspectral imaging and a cutting-edge precise microscopic optical imaging technology such as CLE. This confirms the potential of HSI in identifying marginally perfused areas of gastric conduit tract and the usefulness of HYPER as a potential surgical navigation tool for this type of surgery. Despite being promising, our findings must be confirmed in the clinical setting with lager numbers in order to be able to draw any clinical and cost/benefit conclusion.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.