COMPARATIVE EVALUATION OF ENDOSCOPIC FACTORS FROM CONVENTIONAL COLONOSCOPY AND NARROW-BAND IMAGING OF COLORECTAL LESIONS

Abstract

Background:  Narrow-band imaging (NBI) diagnosis of colorectal lesions requires further consideration and standardized classification for attaining consensus among endoscopists.

Objective:  Compare vascular findings and diagnosis of colorectal lesions using various endoscopic images to assess effectiveness of NBI with magnification.

Method:  Three endoscopic factors (vascular findings; diagnosis from NBI with magnification images; and total diagnosis from all available colonoscopic images including white light endoscopy, chromoendoscopy, NBI with magnification and magnifying chromoendoscopy) for 15 colorectal lesions were evaluated by 12 expert colonoscopists (Group A), 36 endoscopists not expert in colonoscopy (Group B) and 55 individuals not medical doctors, but familiar with gastrointestinal endoscopy or other medical products (Group C). We calculated accuracy percentages for all three endoscopic factors based on responses from each group. Highest percentage for vascular finding, NBI with magnification diagnosis and total diagnosis for each lesion assumed to be opinion of each group. We compared incidence of agreement among three groups and calculated invasion depth diagnostic accuracy rate for each group.

Results:  For vascular findings from NBI with magnification images, incidence of complete agreement among three groups was 46.7%, incidence of disagreement among three groups was 46.7% and incidence of complete disagreement among three groups was 6.7%. For diagnosis from NBI with magnification images and total diagnosis from all available endoscopic images, incidence of complete agreement, disagreement and complete disagreement were 40.0%, 60.0% and 0.0%, respectively, and 60.0%, 40.0% and 0.0%, respectively. Invasion depth diagnostic accuracy from NBI with magnification images and all available endoscopic images were 60.0% and 73.3%, respectively, in Group A; 46.7% and 66.7%, respectively, in Group B; and 46.6% and 60.0%, respectively, in Group C.

Conclusion:  Results suggest some level of educational training is required for effective application of NBI with magnification for precise diagnosis of colorectal lesions.

INTRODUCTION

Narrow-band imaging (NBI) is a recently developed method of endoscopic detection and diagnosis that has been under study for a relatively short period. There are currently several reported classifications for NBI diagnosis of colorectal lesions in Japan^1–4 and additional time will be required for attaining a consensus among endoscopists. Consequently, NBI diagnosis requires further consideration and a standardized classification before achieving such a consensus. We conducted a comparative evaluation of vascular findings and diagnosis of colorectal lesions using conventional endoscopy and NBI with magnification for the purpose of assessing the effectiveness of NBI with magnification.

METHODS

Endoscopic images of lesions

Fifteen colorectal lesions (three adenomas, six intramucosal cancers and six submucosal cancers including one superficial submucosal cancer (<1,000 µm) and five deep submucosal cancers (≥1,000 µm) that had been examined from October 2005 to May 2010 using NBI with magnification, conventional white light endoscopy, chromoendoscopy and magnifying chromoendoscopy were selected for this study. Surgical or endoscopic resections were carried out on all 15 lesions and histopathological diagnoses were made by a highly experienced pathologist. The best quality NBI with magnification, white light endoscopy, chromoendoscopy and magnifying chromoendoscopy images for each lesion were digitally stored in JPEG format. (Fig. 1)

Figure 1. Endoscopic images used in this study. (a) Narrow-band imaging with magnification images. (b) White light endoscopy, chromoendoscopy and magnifying chromoendoscopy images.

Image evaluators

A total of 103 individuals participated as image evaluators in this study including 12 expert colonoscopists (Group A), 36 endoscopists who were not experts in colonoscopy (Group B) and 55 individuals who were not medical doctors (Group C). None of the Group C evaluators had been educated or trained in endoscopic diagnosis before this study although all of them were engineers or sales agents for gastrointestinal endoscopy or other medical products. These individuals were included in the study to show whether it was possible to diagnose capillary pattern irregularities and malignancies based solely on their observational abilities without any previous endoscopic diagnosis education or training.

Evaluation factors

Three endoscopic factors (vascular findings based only on NBI with magnification images; diagnosis based only on NBI with magnification images; and total diagnosis based on all available endoscopic images (white light endoscopy, chromoendoscopy, NBI with magnification and magnifying chromoendoscopy)) were evaluated in this study. Vascular findings from only NBI with magnification images were separated into four categories (regular network, irregular network, coarse vessels and fragmentation, and avascular area). Diagnoses from only NBI with magnification images as well as total diagnoses from all available endoscopic images were divided into three types (adenomas, intramucosal and superficial submucosal cancers [<1,000 µm], and deep submucosal cancers [≥1,000 µm]). (Table 1)

Table 1.  Image evaluation questionnaire/answer sheet

NBI, narrow-band imaging, M-SM, mucosal-submucosal; SM, submucosal.

Q1: What vascular findings did you choose in this NBI magnifying image?

1. Regular network

2. Irregular network

3. Coarse vessels & fragmentation

4. Avascular area

Q2: How do you diagnosis from this NBI image?

1. Adenoma

2. M-SM slight invasion

3. SM massive invasion

Q3: How do you diagnosis from total image?

1. Adenoma

2. M-SM slight invasion

3. SM massive invasion

Assessment of endoscopic images

First, the evaluators examined NBI with magnification images of the 15 colorectal lesions and then selected the best description for each lesion from the four possible vascular finding categories (regular network, irregular network, coarse vessels and fragmentation, and avascular area). Next, the evaluators diagnosed each lesion from the three possible types (adenomas, intramucosal and superficial submucosal cancers [<1,000 µm], and deep submucosal cancers [≥1,000 µm]) based only on the NBI with magnification images. Finally, the evaluators examined white light endoscopy, chromoendoscopy and magnifying chromoendoscopy images of the 15 colorectal lesions and then made a total diagnosis from the three possible types for each lesion taking into account all endoscopic images. Evaluators made diagnostic decisions in real time using response cards so as to answer immediately and were subsequently prohibited from changing an initial NBI diagnosis after making a total diagnosis for a particular lesion. Patient information such as age, sex and clinical diagnosis was not disclosed to any of the evaluators and discussions were not permitted among the 48 endoscopists either individually or in groups.

Statistical analysis

We calculated the percentages for all three endoscopic factors based on the responses from each group. The highest percentage for the vascular finding, NBI with magnification diagnosis and total diagnosis for each lesion was assumed to be the opinion of each group. We then compared the percentages for the three groups for each of the three endoscopic factors. ‘Complete agreement’ meant all three groups made the same choice. ‘Disagreement’ meant one of the groups made a choice that was different from the other two groups whereas ‘complete disagreement’ meant there were three completely different choices made by the three groups. We also calculated the invasion depth diagnostic accuracy rate for each group. All statistical analyses were carried out using a Microsoft Excel 2007 spreadsheet (Microsoft, Renton, WA, USA).

RESULTS

Comparative agreement on vascular findings and diagnoses

For vascular findings of the 15 colorectal lesions using only NBI with magnification images, the incidence of complete agreement among the three groups was 46.7%, the incidence of disagreement among the three groups was 46.7% and the incidence of complete disagreement among the three groups was 6.7%. For diagnosis using only NBI with magnification images and total diagnosis using all available endoscopic images, the incidence of complete agreement, disagreement and complete disagreement among the three groups were 40.0%, 60.0%, 0.0%, respectively, and 60.0%, 40.0%, 0.0%, respectively. (Table 2)

Table 2.  Comparison of assessment agreement rates for three endoscopic factors

	Vascular findings – only narrow-band imaging (%)	Diagnosis – only narrow-band imaging (%)	Total diagnosis – all endoscopy images (%)
Complete agreement	46.7	40.0	60.0
Disagreement	46.7	60.0	40.0
Complete disagreement	6.7	0.0	0.0

The incidence of agreement between Group A/Group B, Group B/Group C and Group A/Group C on vascular findings using only NBI with magnification images was 66.7%, 60.0% and 60.0%, respectively. For diagnosis using only NBI with magnification images, the incidence of agreement between Group A/Group B, Group B/Group C and Group A/Group C was 73.3%, 60.0% and 53.3%, respectively, whereas the incidence of agreement between Group A/Group B, Group B/Group C and Group A/Group C on total diagnosis using all available endoscopic images was 80.0%, 80.0% and 73.3%, respectively. (Table 3) One case of complete agreement and another case of complete disagreement between the three groups are shown in Figs 2 and 3, respectively.

Table 3.  Comparison of assessment agreement rates between image evaluator groups

	Vascular findings – only narrow-band imaging (%)			Diagnosis – only narrow-band imaging (%)			Total diagnosis – all endoscopy images (%)
	Group A vs. Group B	Group B vs. Group C	Group A vs. Group C	Group A vs. Group B	Group B vs. Group C	Group A vs. Group C	Group A vs. Group B	Group B vs. Group C	Group A vs. Group C
Group A: 12 expert colonoscopists. Group B: 36 endoscopists non-expert in colonoscopy. Group C: 55 individuals who were not medical doctors.
Agreement	66.7	60.0	60.0	73.3	60.0	53.3	80.0	80.0	73.3
Disagreement	33.3	40.0	40.0	26.7	40.0	46.7	20.0	20.0	26.7

Figure 2. Complete agreement among three image evaluator groups on 25 mm polypoid is lesion located in rectum. Histopathological analysis of endoscopically resected specimen showed intramucosal cancer. (a) Narrow-band imaging (NBI) with magnification images. (b) White light endoscopy, chromoendoscopy and magnifying chromoendoscopy images. (c) Detailed questionnaire results for three endoscopic factors. () Expert in CF; () non-expert; () audience. M-SM, mucosal-submucosal; SM, submucosal.

Figure 3. Complete disagreement among three image evaluator groups on 15 mm flat IIa+IIc lesion located in rectum. Histopathological analysis of surgically resected specimen showed deep submucosal cancer (≥1,000 µm). (a) Narrow-band imaging (NBI) with magnification images. (b) White light endoscopy, chromoendoscopy and magnifying chromoendoscopy images. (c) Detailed questionnaire results for three endoscopic factors. () Expert in CF; () non-expert; () audience. M-SM, mucosal-submucosal; SM, submucosal.

For diagnosis from NBI with magnification images and total diagnosis from all available endoscopic images, incidence of complete agreement, disagreement and complete disagreement were 40.0%, 60.0% and 0.0%, respectively, and 60.0%, 40.0% and 0.0%, respectively. Invasion depth diagnostic accuracy from NBI with magnification images and all available endoscopic images were 60.0% and 73.3%, respectively, in Group A; 46.7% and 66.7%, respectively, in Group B; and 46.6% and 60.0%, respectively, in Group C.

Invasion depth diagnostic accuracy

Invasion depth diagnostic accuracy using only NBI with magnification images and using all available endoscopic images was 60.0% and 73.3%, respectively, in Group A; 46.7% and 66.7%, respectively, in Group B; and 46.6% and 60.0%, respectively, in Group C (Table 4) although there were no significant differences among the three groups in invasion depth diagnostic accuracy using only NBI with magnification images or using all available endoscopic images.

Table 4.  Comparison of invasion depth diagnostic accuracy rates

	Diagnosis – only narrow-band imaging (%)	Total diagnosis – all endoscopy images (%)
Group A	60.0	73.3
Group B	46.7	66.7
Group C	46.6	60.0

DISCUSSION

Total diagnosis in Group A had the highest diagnostic accuracy followed by total diagnosis in Group B. NBI with magnification diagnostic accuracy in Group A was lower than total diagnosis in Group B and the same as total diagnosis in Group C. NBI with magnification diagnosis in Group B and Group C produced the lowest diagnostic accuracy rates recorded in this study. These results suggest that some level of educational training is required for the effective application of NBI with magnification for precise diagnosis of colorectal lesions.

There are currently several reported classifications for NBI diagnosis of colorectal lesions^1–4 so it might be some time before a consensus is reached among endoscopists on NBI with magnification diagnosis in comparison to pit pattern diagnosis and conventional endoscopic diagnosis. Throughout the floor discussion, it was evident that agreement on determining regular from irregular capillary patterns was an unresolved issue even among expert colonoscopists thereby indicating that NBI diagnosis will require further consideration and a standardized classification must be established before a consensus is possible.

Over the last 20 years, pit pattern analysis has resulted in precise endoscopic diagnosis including differential diagnosis among hyperplastic polyps, adenomas and cancers in addition to accurate depth of invasion diagnosis for early colorectal cancers.^5–7 Pit pattern diagnosis is now considered more objective than conventional endoscopic diagnosis although it still remains subjective in diagnosing pit pattern irregularities.

In contrast, NBI is a recently developed method of endoscopic detection and diagnosis that has been under study for a relatively short period. NBI with magnification facilitates detailed observation of a lesion's microvascular architecture and is widely used for the endoscopic diagnosis of colorectal tumors. Although NBI with magnification has proven very useful for differential diagnosis between non-neoplastic and neoplastic lesions using capillary pattern classification,^8–10 there is no consensus as yet on depth of invasion diagnosis or determining the borderline between regular and irregular networks despite some agreement on the definition of capillary irregularity.

Based on the floor discussion, we realize that our most urgent task now is to standardize a classification for diagnosis using NBI colonoscopy. A multicenter study will then be required to evaluate the diagnostic accuracy of NBI with magnification in determining invasion depth compared with the existing conventional endoscopic classification in order to build a consensus. Undoubtedly, it will be difficult to exclude interobserver variability from such an NBI classification for colorectal lesions so interpretation of NBI with magnification findings based on an objective set of standards will also be necessary. Computer-assisted analysis might be one means of reducing the subjective nature of endoscopic diagnosis. Once achieved, a highly accurate automatic identification process should substantially enhance the educational training of less-experienced endoscopists in the use of NBI with magnification for the diagnosis of colorectal lesions.