1. Introduction

Gastrointestinal carcinomas have the highest incidence (19.2% of all new cancer cases) and mortality worldwide (23% of annual worldwide cancer mortality, 1.81 million [1]). As a result of an increase in endoscopic (screening) examinations and the continuous improvement of the optical quality of endoscopic equipment, T1 carcinomas in the gastrointestinal tract (GIT) are increasingly detected and diagnosed. e.g., 7–17% of colonic T1 carcinomas already have LM [2], which significantly depends on the depth of invasion. This number was significantly lower (11%) in the group of correctly characterized T1 carcinomas.

However, with aids such as attachment caps, underwater magnification and digital zoom, magnifications of up to 75-fold can be achieved with conventional HD endoscopes. Complemented by contrast-enhanced endoscopy (image-enhanced endoscopy, IEE) based on staining (true chromoendoscopy, CE) or device-based, virtual chromoendoscopy (e.g., as narrow-band imaging, NBI or blue light imaging, BLI), the accuracy of endoscopic characterization of lesions in the GIT could be improved dramatically. Device-based IEE, such as NBI, enables improved contrasting of vascular structures by emission of blue and green light of a defined wavelength (in the absorption range of hemoglobin). Based on changes in the microsurface structure (SP) and the microvascular pattern (capillary structures, VP), numerous classifications of pathological changes have been published in the various areas of the GIT in recent years, the most important of which will be presented and evaluated in terms of evidence and practical application in this review.

2. Examination Procedure, Detection and Characterization

A high quality of the endoscopic examination (e.g., good bowel preparation, mucolytic/surface preparation of the mucosa, meticulous examination technique, standardized examination protocols) is the prerequisite for reliable detection and characterization of suspicious lesions Already in the first step, on the basis of morphology and certain additional features (e.g., sharp delineation, inhomogeneity of color distribution), reliable conclusions about the endoscopic resectability of the lesion can often be made, not requiring ME (NICE classification in the colon and rectum ^[3][4][6,7], GUP system in the stomach ^[5][8]). Thus, more than 50% of LST (lateral spreading tumors) of the non-granular type in the colon and up to 80% of all flat (0-II) lesions in the stomach are associated with malignant histology ^[6][7][11,12]. To further characterize gastrointestinal neoplasms (dignity, lateral extent, and depth of invasion), VP and (ideally with CE)

2.1. Esophagus

In the squamous esophagus, sessile/polypoid (0-Is/p) or ulcerative neoplasms (0-III) have the highest risk of deep submucosal invasion (SMI, >80%). However, even shallow lesions (0-II) already have a risk of SMI of 15–53% [37]. Similarly, in Barrett’s esophagus, up to 45% of 0-IIa lesions already represent T1b carcinomas [38]. Therefore, a closer characterization at the micro-level by combining ME and IEE is required. Only two studies in the context of squamous cell carcinoma (SCC) met the inclusion criteria (Table 1). The recently published JES classification of the Japanese Esophageal Society, which is a further development of the IPCL classification [39], showed a statistical accuracy of 91% regarding the depth of invasion in a prospective study [15]. This was also confirmed in a recent meta-analysis [14]. This classification is based on the pattern of microvessels (VP) and their degree of dilatation. Four VP types are defined, which allows for a subtle assessment of the degree of submucosal invasion of SCC in the esophagus. Only intramucosal SCC is appropriate for endoscopic resection (JES type B1). Lesions invading the muscularis mucosae or the upper layer of the submucosa (JES type B2) are a relative contraindication for endoscopic resection.

Also in Barrett’s esophagus, ME-NBI (for VP) and acetic acid-based ME-CE (for SP) can identify areas of high-grade intraepithelial neoplasia or early carcinoma with more than 90% sensitivity and specificity, and low interobserver variability ^{[10][11][29][30]}[16,17,40,41]. Only two studies in the context of Barrett’s carcinoma met the inclusion criteria (BING classification validation study and simplified NBI classification, Table 1). The most recent of the existing classifications of Barrett’s esophagus is the JES-BE classification, which applies the clinically well-studied criteria in gastric carcinoma to the changes in Barrett’s carcinoma [42]. In brief, similarly to the BING classification, mucosal and vascular patterns are classified separately (regular or irregular). In the JES-BE classification, visibility and morphologic features (e.g., pit or network) are specified additionally. This classification reflects the natural highly variable appearance of Barrett’s lesions. Of note, this classification has not yet been validated in clinical prospective studies. However, quadrant biopsies according to the Seattle protocol cannot be dispensed, as about half of early malignant neoplasms are detected by “protocol biopsies” ^[31][32][]. Assessment of depth of invasion is challenging in Barrett’s (early) carcinoma—in practice, a combined view of morphology and SP/VP is recommended. Deep SMI is very likely in all sessile (0-Is), elevated (0-IIa), sunken (0-IIc), or ulcerated (0-III) lesions in combination with a highly irregular VP (i.e., high variability of vessel diameters, low density) and/or amorphous or very irregular SP [35].

2.2. Stomach

In order to achieve the highest detection rates and accuracy of characterization, a combination of meticulous WLE followed by ME-NBI is currently recommended [44]. Already with WLE, a high quality of detection of suspicious lesions can be achieved after good preparation, taking into account the medical history (risk factors such as helicobacter pylori infection). Here, a standardized systematic screening protocol is recommended [45]. Localization, color, demarcability to the surrounding mucosa, and morphology should be analyzed for each suspicious lesion. However, beyond WLE, the ME-NBI technique which allows analysis of the specific anatomical structures (e.g., subepithelial capillaries and marginal crypt epithelium), especially in small gastritis-like lesions, is crucial for characterization. Eight studies and meta-analyses met the inclusion criteria (Table 1), showing a very high diagnostic accuracy (>95%) of ME-NBI in characterization in the context of early gastric cancer. Figures 3 and 4 illustrate this diagnostic process and the correlations with histologic criteria (differentiated early carcinoma in the antrum). The “vessel plus surface” classification enables differential diagnosis of differentiated early carcinoma (sharp demarcation in combination with irregular SP and/or VP) with a sensitivity and specificity of 95%, which was confirmed in a recently published meta-analysis [20]. In this classification, three categories exist for SP and VP, respectively: regular, irregular, and absent. The presence of an irregular VP with a demarcation line, or the presence of an irregular SP with a demarcation line is diagnostic for early gastric cancer. In this context, the “white zone” is equivalent to the gastric SP, anatomically correlated to the marginal crypt epithelium. In Figures 3 and 4, all features of the vessel plus surface classification are shown for a neoplastic lesion and normal mucosa with corresponding histopathological sections, respectively. Weaknesses exist in the diagnosis of undifferentiated gastric carcinomas (G3, G4), which may appear as pale, sunken lesions on WLE, and may often have poorly defined delineation and, in some circumstances, an inconspicuous SP (if diffusely subepithelial) on ME-NBI. The depth of invasion of differentiated carcinomas is reliably assessed by macroscopic morphology (Paris type, fold criteria), VP, and SP. Thus, a raised or sunken/ulcerated lesion with loss of SP has a >80% risk of massive SMI (≥SM2, [46]).

2.3. Colon

Nine prospective studies and meta-analyses met the inclusion criteria (Table 1), showing a very high diagnostic accuracy (>90%) of ME-NBI in characterization in the context of early colorectal cancer. To achieve >90% statistical confidence in the characterization of colorectal neoplasms it is essential to evaluate colorectal neoplasms according to VP (Sano or JNET, in ME-NBI) and SP (pit pattern according to Kudo, in ME-CE [32,47][32,46]). In Figure 5, two examples of endoscopic diagnosis (WLE, NBI, and CE) of colorectal lesions are demonstrated. Both lesions show a centrally sunken surface when examined by WLE alone (IIa + IIc). ME-NBI shows an irregular dense or highly irregular rarefied vascular pattern (VP) in the central area, respectively. Supplementary CE allows for the assessment of the VP (irregular, high density, or amorphous surface). JNET classification is based on this analysis algorithm to distinguish hyperplastic, adenomatous, superficial malignant, and deeply invasive lesions [48]. In brief, for each lesion VP and SP are evaluated separately. Variable caliber or irregular distribution of VP and/or irregular or obscure SP are diagnostic for high-grade intramucosal neoplasia or shallow submucosal cancer (JNET Type 2B). Loose vessel areas or interruption of thick vessels and/or amorphous SP are diagnostic for deep submucosal invasive cancer (JNET Type 3), not suitable for endoscopic resection. Morphological criteria such as nodular formations on polyps, amorphous color changes, bleeding tendency, and central ulceration with lack of surface structure, as well as fixed deformities after desufflation have high importance in terms of estimation of deep SMI (>1000µm) and endoscopic non-resectability. In this regard, analysis by ME-NBI supported by CE (crystal violet, indigo carmine) could show sensitivities of 84–90% and a pooled specificity of 98% in a recent meta-analysis (independent of the classification system used). This is significantly superior to a mere WLE diagnosis [27,33]. The NBI international colorectal endoscopic classification (NICE), which utilizes NBI without ME, allowed for a specificity of >96% in identifying invasive, i.e., endoscopically unresectable T1 colorectal tumors [31].

3. Discussion