Differences in the Microsurface and Microvascular Features between Early Gastric Cancer of Corpus and Antrum under Narrow Band Imaging Magnifying Endoscopy

Microsurface structure (MS) and microvascular features (MV) of early gastric cancer (EGC) are associated with the histologic type, invasion depth and macroscopic shape of lesions. However, no studies have reported the differences in the MS and MV features between EGCs of antrum and corpus under magnifying endoscopy with narrow band imaging (ME-NBI).Methods This was a retrospective study including 116 patients with EGC. The characteristics of MS and MV pattern of EGC were evaluated in 2 group, antrum type and corpus type.Results the of undifferentiated-type was times high the differentiated-type No difference in the MV pattern was between antral and corporal significantly different between those 2 groups (P=0.001).Conclusions


Background
Magnifying endoscopy with narrow band imaging (ME-NBI) is able to observe the features of microsurface and microvessels in the gastric mucosa. Studies have confirmed the value of ME-NBI in the diagnosis of early gastric cancer (EGC) [1][2][3][4].
Prior studies have demonstrated that ME-NBI can achieve endoscopic pathology of the upper gastrointestinal tract, in which an accurate endoscopic diagnosis comparable to the histopathological diagnosis can be made without biopsy [1,[5][6][7].
That is to say, ME-NBI can provide very clear endoscopic images in which critical histopathological features (such as structural atypia) can be identified [8]. It has also been proven to be able to predict the histological characteristics of gastric cancer lesions, such as the degree of differentiation [9], the depth of infiltration [10,11], etc. ME-NBI can also assist to determine the tumor boundaries and avoid positive postoperative surgical margins [12]. Therefore, it is an effective method for preoperative evaluation of tumors.
Microsurface structure (MS) and microvascular features (MV) of EGC are associated with the macroscopic shape, histologic type and invasion depth of lesions [13]. It is well known that there are significant differences in MV and MS between the mucosa of normal antrum and corpus under ME-NBI [4,14]. Normal corporal mucosa displays a regular honeycomb-like subepithelial capillary network (SECN) and regular oval crypt opening with circular marginal crypt epithelium (MCE) [14]. Normal antral mucosa appears as a regular coil-shaped SECN pattern and regular polygonal or curved MCE pattern [14]. However, no studies have reported the difference between EGCs of the corpus and antrum under ME-NBI. The MV and MS features of EGC summarized in previous studies also cannot be used to distinguish between the different regions of antrum and corpus [13,[15][16][17][18]. We hypothesize that there may be different MV and MS patterns between EGCs of the corpus and antrum under ME-NBI.

Study design and patient selection
This retrospective study was carried out at a single tertiary referral center in Shanghai China in conformity to the declaration of Helsinki, and was approved by the institutional review board of Shanghai General Hospital of China.
The patients meeting the following inclusion and exclusion criteria were enrolled into the present study from Shanghai General Hospital between September 2014 and March 2019. ME-NBI was conducted before treatments and the images were stored in a digital filing system. All patients underwent endoscopic submucosal resection or surgical treatment and were histologically diagnosed with EGC. Lesions with inappropriate NBI images due to bleeding, ulcer covered by corpus albicans, or excessive mucosal congestion were excluded from the study. Finally, a total of 116 EGCs from 116 patients were included in the present study.

ME-NBI
ME-NBI was performed with a magnifying endoscope (GIFH260Z, Olympus Co., Ltd, Japan) (GIFH260Z, Olympus, Tokyo, Japan) equipped with a video processor (EVIS LUCERA Olympus CV-290, Olympus Co. Ltd), and a light source (EVIS LUCERA Olympus CLV-290SL, Olympus co.). The structure enhancement of the endoscopic video processor was set at B-mode level 8 for ME-NBI. The color mode was fixed at level 1. To obtain stable endoscopic images at maximal magnification, a black, soft hood attachment (MAJ-1990, Olympus co.) was mounted at the tip of the endoscope before examination.
The MS and MV of EGC areas were retrospectively evaluated using ME-NBI images captured at the maximal magnification by an experienced endoscopist who was blind to the histological findings.
Studies have reported that epithelial structures can be observed on ME-NBI [2,6,9]. In the present study, structural atypia was used to describe the degree of microsurface structure disorder. According to the Sakaki types [19], the magnifying images of EGC areas were classified into three types [20]: IV, V 1 and VI based on MS patterns. Type IV is ovoid or patchy and type V 1 is finger or villous structure. The microsurface structure of type VI is often disordered, destroyed or even absent. As the increase in the classification level, the microsurface structure becomes more disordered and is more prone to be destroyed and absent or in other words the grade of atypia of microsurface structure gets higher (Fig. 1). The MV patterns were classified as loop, fine network, or corkscrew according to the classification previously reported [16] (Fig. 2).
According to the classification proposed by Yokoyama et al [13], NBI magnifying images were also classified into microvascular pattern and microsurface pattern (VS patterns): FNP, ILL-1, ILL-2 and CSP, which were simultaneously based on the abnormal VS patterns (Fig. 3). The predominant pattern was used for classification in case of lesions with combined MS or MV patterns.

Histological assessment
The removed tumors were fixed in 10% buffered formalin for 24h and then longitudinally cut into 2-3 mm slices. These sliced were embedded in paraffin and stained with hematoxylin and eosin for pathologic examination. In the present study, the revised Vienna [21] categories 4 and 5 were used to define EGC. The depth of invasion and degree of differentiation were assessed microscopically. The histologic diagnosis of EGC was according to WHO classification [22]. The WHO classification is commonly divided into 2 major categories in clinical study: differentiated and undifferentiated type (Nakamura's classification) [23,24]. The former includes well-differentiated adenocarcinoma, moderately-differentiated adenocarcinoma, and papillary adenocarcinoma, and the latter includes poorlydifferentiated adenocarcinoma, signet-ring cell carcinoma, and mucinous adenocarcinoma. In this study, we utilized this classification method to divide EGC into two types: differentiated and undifferentiated type.

Statistical analysis
Because the Sakaki types of MS patterns were ordered categories, Wilcoxon ranksum test was used in univariate analysis, and ordinal ploytomous logistic regression was used in multivariate analysis, so as to compare the degree of atypia of MS between different groups. However, unlike MS patterns, the types of MV patterns and VS patterns were unordered categories and their distribution between different groups was analyzed using the Fisher exact test. Statistical analyses were performed with IBM SPSS Statistics for Windows (Version 24.0) and the free software R (www.r-project.org). A value of P <0.05 was considered statistically significant.

Participants
In total, 116 EGC lesions were examined in the present study. Thirty-nine lesions located in the corpus and 77 in the antrum. Fifty-six lesions were treated via ESD and 60 via surgery. The results of histologic assessment based on histologic type and invasion depth were as follows: 70 lesions were differentiated-type EGCs (D-EGCs) and 46 undifferentiated-type EGCs (U-EGCs); 95 lesions were mucosal cancers and 21 submucosal cancers (Table 1). Data are presented as number (%) or mean (range).
According to the lesion shape reported by Yokoyama et al [13], the lesions were classified into 2 groups according to their morphology: elevated-type EGC (type 0-I, type IIa, and type IIa+IIc; predominantly IIa); and flat-and depressed-type EGC that consisted of type IIb and type IIc. As a result, 49 lesions were classified as elevatedtype EGC and 67 lesions as flat-and depressed-type EGC.

MS patterns and tumor location of EGCs
As shown in Table 2, tumor location (P<0.001), histologic type (P<0.001), invasion depth (P=0.002) and macroscopic shape (P=0.03) were significantly associated with the types of MS patterns (the degree of structural atypia) in univariate analysis. In  Fig. 4-6) and U-EGCs was greater than D-EGCs.

MV patterns, VS patterns and tumor location of EGCs
It was reported that MV patterns and VS patterns were associated with the histologic type, invasion depth and macroscopic shape of lesions [13], which were confounding factors when the location factor was assessed. Before Case Control Matching, the distribution of invasion depth was significantly associated with the tumor location of EGCs (c 2 =8.0, P=0.005) (

MS patterns and tumor location of EGCs
Although the number of cases got smaller after case control matching 1:1, MS patterns were still significantly different between corpus group and antrum group (P=0.009), which was consistent with the above result of ordinal ploytomous logistic regression (Table 5).

MV patterns and tumor location of EGCs
As shown in Table 6, there was no difference in the MV patterns of EGCs between antrum group and corpus group (P=0.7).

VS patterns and tumor location of EGCs
As shown in Table 7, the classification categories were significantly different between corpus group and antrum group (P=0.001). The proportion of FNP in the corpus was higher than in antrum group, while the proportion of ILL-1 in antrum group was much higher than corpus group. There was no significant difference in the distribution of ILL-2 and CSP between those two groups.

Discussion
In the present study, results showed that the MS patterns of EGC on me-nbi were significantly different in the gastric antrum and the corpus. Compared with the antral lesions, the grade of atypia in MS patterns was higher in corpus group. Based on the VS classification system, antral lesions were more likely to be ILL-1 type, while the corporal lesions were more inclined to be FNP type in those 2 groups of EGC.
In this study, the gastric pit pattern was evaluated according to the system proposed by Sakaki et al [19]. Although there are some limitations in this system, currently there is no other internationally recognized classification system for the assessment of gastric pits on upper digestive tract magnifying endoscopy.
Therefore, this classification is still widely used in clinical practice [25]. According to the classification, the types of gastric pit pattern of EGC or precancerous gastric lesions mainly presented as IV, V 1 or VI type [20]. In the present study, we found that the grade of atypia in MS patterns in the corpus type was more likely to be higher than in the antrum. In other words, the disordered or missing MS patterns in the corpus group were more obvious than the antrum group (Fig. 4-6).
That's an interesting finding. It is known that structural atypia and cellular atypia are key histopathological features for cancer diagnoses. ME-NBI can produce optical images reflecting structural atypia, which can achieve "optical biopsy" [8]. The structural atypia of gastric gland (or MS) of EGC may present as disordered or absent under ME-NBI. In that way, what's the reason that the degree of MS atypia in the corpus group was significantly higher than that in the antrum group?
We speculate that this phenomenon could be explained by the structure of gastric wall mucosa, the occurrence of tumor epithelium, and the characteristics of tissue atypia. Histologically, the gastric pit comprises approximately 50% of antral mucosal thickness, but only 25% of corporal mucosal thickness. There is evidence showing that gastric stem cells usually reside in the base of gastric unit and gastric cancer stem cells (GCSCs) play an important role in the occurrence and development of gastric cancer [26]. Thus, we assume that the proliferation of GCSCs may cause gradual glandular morphological changes which arise from the bottom.
The superficial glandular structure in the corpus is more susceptible to these changes because the crypts in the corporal mucosa are shallower. When the gastric mucosa becomes cancerous, corporal microsurface is prone to be disorganized, which can be easily observed on ME-NBI.
Previous researches have also proposed that the degree of differentiation of tumor tissue affects the appearance of the microsurface [9, 27]. Kanesaka et al [9] reported that the absence of MS patterns was the characteristic of U-EGCs because cancer cells in U-EGC usually failed to form glands and the original epithelial structure was often destroyed, while cancer cells in D-EGCs usually had epithelial glandular structures, which showed subtle structural atypia. Our study also suggested the structural atypia of MS patterns in U-EGCs was significantly different from that in D-EGCs, which was consistent with the pathological features of those 2 types of cancers.
However, the relationship between histological type factors and magnifying images may also be influenced by other common factors, such as depth of invasion [10], macroscopic morphology [13], and tumor location. Similarly, the relationship between tumor location and microstructure of EGC under me-nbi was also affected by other confounding factors, including depth of invasion and macroscopic morphology, in addition to histological type. In order to comprehensively analyzed whether those four factors were correlated with MS patterns, a common statistical method for multivariate analysis was employed to assess the whole cases in this study. The analysis confirmed that only tumor location and histologic type were independent predictors for MS patterns ( Table 3).
As a result, in the EGCs, compared with the corpus, the atypia of MS pattern EGCs in the antrum were more subtle and thus their diagnosis might be more likely to be missed clinically. In daily endoscopic examination, there are usually multiple lesions coexisting in the gastric antrum, while EGC may only show slight changes in microstructure and microvascular patterns [28]. EGCs in the gastric antrum area are easily missed or difficult to distinguish from non-cancerous lesions due to the above reasons. Therefore, when equivocal lesions are found in the antrum, additional biopsies may be essential for inexperienced endoscopist.
Because the MV patterns and VS patterns were associated with the histologic type, invasion depth and macroscopic shape of lesions [9, 13], case control matching was employed to adjust these confounding factors before further analysis of tumor location. As a result, according to the MV classification used in this study, there was no significant difference in MV patterns between EGCs in the antrum and corpus.
The microvessels in the superficial cancerous epithelium are deformed as a result of compression by the structurally atypical cancerous glands [29]. Thus, the secondary changes may not be different between the antrum type and the corpus type.
Yokoyama et al [13] reported that the distribution of VS patterns was significantly different between D-EGCs and U-EGCs. In this study, the classification categories between D-EGCs and U-EGCs were consistent with those reported by Yokoyama et al (Additional file 1: Table S1). As a result, the distribution of VS patterns was also significantly different between the corpus type and the antrum type of EGC. As we know, normal corporal mucosa appears as a regular honeycomb-like SECN pattern and regular oval crypt opening with circular MCE pattern [14]. Normal antral mucosa displays a regular coil-shaped SECN pattern and regular polygonal or curved MCE pattern [14]. FNP and ILL-1 are predominantly found in differentiated-type adenocarcinomas [13], in which there are still epithelial glandular structures with subtle structural atypia. Morphologically, FNP is similar to the normal corpus MS pattern and ILL-1 to the normal antral MS pattern. Correspondingly, in EGC lesions, FNP is mainly observed in the corpus type, while ILL-1 is more common in the antrum type.
However, the incidence of ILL-2 and CSP in the gastric antrum and gastric corpus is comparable. Based on previous work [13] and our findings (Additional file 1: Table   S1), ILL-2 and CSP are mainly found in undifferentiated adenocarcinoma. We hypothesize that, since their glandular and epithelial structures are often destroyed in U-EGCs, the pattern of microstructure may not differ greatly between the antrum type and the corpus type. In clinical practice, these findings are helpful for the better understanding of the likelihood of these four VS categories in different sites and degrees of differentiation of EGCs.
There are also some limitations in this study. First, this was a retrospective study from a single center, and the sample size was relatively small. More multicenter, prospective studies are needed to confirm our findings. Second, no significant difference was observed between the MV patterns of EGCs in the antrum and corpus in the present study, according to the current MV classification. Whether this is related to the method used for classification of MV patterns is unclear.

Conclusions
In  Microsurface pattern in EGCs on ME-NBI.

Figure 2
Microvascular pattern in EGCs on ME-NBI. Comparison between the cases of high-grade intraepithelial neoplasia (HGIN) in corpus and a  Additional file 1.xlsx