White-light endoscopic imaging

Standard videoendoscopy magnifies an image approximately 5 to 10 times and gives an image resolution of 100,000 to 300,000 pixels. A pixel is a photosensitive element that generates an electrical charge in proportion to light exposure and then generates an analog signal that is digitalized by a computer video processor. Image resolution (ie, the ability to discriminate between 2 adjacent points) achieved with a standard video endoscope is not sufficient, as the focal distance of these endoscopes is only 1 to 9 cm. This distance restricts the ability of standard white-light endoscopes to detect dysplasia and early carcinoma with precision, as any lesion beyond the range of this focal distance will not appear clear. Hence, quadratic mucosal biopsies are required to make the diagnosis. However, biopsy sampling is associated with certain issues such as increased cost, more time required, more sampling errors, and high interobserver variability in histopathological diagnosis. Techniques are needed that can easily differentiate normal from abnormal mucosa, to guide and, if possible, eliminate the need for biopsies. High-resolution endoscopes (HRE) are a newer class of endoscope that better detect microscopic mucosal abnormalities because of their ability to produce images with higher magnification (of up to 100 times) and increased spatial resolution of 600,000 to 1 million pixels. HRE are equipped with movable lenses, resulting in variable focal distance. These qualities allow more detailed examination of the mucosal, glandular, and vascular structures at a range of less than 3 mm.

Chromoendoscopy

Chromoendoscopy involves endoscopic examination of gastrointestinal mucosa after applying a dye solution to the surface, which enhances the appearance of otherwise nonperceivable mucosal changes. This technique guides the endoscopist to take target biopsies only from the suspicious areas. Different dye solutions are used including Lugol iodine solution, methylene blue (MB), indigo carmine, crystal violet, and acetic acid. Lugol solution consists of a 0.5% to 3.0% aqueous solution of potassium iodide, and iodine and is taken up by glycogen-containing cells. It is more useful for detecting dysplasia in patients with increased risk of squamous cell carcinoma (ie, with a history of smoking, alcoholism, and lye ingestion). On the other hand, MB, indigo carmine, and acetic acid are more useful in the detection of glandular abnormalities, as seen in adenocarcinoma. After staining with these agents, different mucosal patterns become visible during videoendoscopy. Various studies have been done to classify and detect dysplasia and early carcinoma based on these patterns. Guelrud and colleagues used acetic acid, a mucolytic agent, to stain the esophageal mucosa, and observed 4 pit patterns (round, reticular, villous, and ridged). Villous and ridged patterns were found to be associated with intestinal metaplasia. Sharma and colleagues studied the mucosal changes in 80 patients with Barrett esophagus (BE) after application of indigo carmine, and observed 3 mucosal patterns: ridged/villous, circular, and irregular/distorted. The irregular/distorted pattern was found to be associated with Barrett high-grade dysplasia (HGD) or superficial adenocarcinoma in 6 patients. In the same study, the presence of the ridged or villous pattern had 97% sensitivity, 76% specificity, and 92% positive predictive value (PPV) for prediction of intestinal metaplasia. However, in subsequent studies, these results could not be reproduced and no detection benefit for either Barrett metaplasia or dysplasia could be established with chromoendoscopy. Other studies using acetic acid and crystal violet stain also produced variable results. The probable reasons for these conflicting observations could be differences in technique, operator experience, and a patient population with a prevalence of BE. This was further substantiated in a blinded European study in which 4 expert gastrointestinal endoscopists analyzed magnification chromoendoscopy images of BE, using acetic acid or MB. The interobserver agreement was poor (κ = 0.40) for all parameters studied, including the mucosal patterns, MB positive staining, and the presence of specialized intestinal metaplasia (SIM).

Ngamruengphong and colleagues performed a meta-analysis of 9 studies published in PubMed for assessment of the diagnostic yield of techniques of chromoendoscopy compared with conventional 4-quadrant random biopsy (RB) in detection of SIM and dysplasia in patients with BE. A total of 450 patients with BE were reported in 9 studies included in the meta-analysis. Data on the yield of both modalities were extracted and analyzed to estimate weighted incremental yield (IY) and 95% confidence intervals (CIs) of MB chromoendoscopy with target biopsies over RB protocol using the Cochrane Q χ ² test. There was no significant IY with MB over RB for detection of SIM (IY 4%; 6 studies, n = 251), dysplasia (IY 9%; 9 studies, n = 450), and HGD or early cancer (EC) (IY 5%; 8 studies, n = 405). This meta-analysis shows that the technique of MB chromoendoscopy only has a comparable yield with white-light endoscopy using RBs for the detection of SIM and dysplasia during endoscopic evaluation of patients with BE.

Ormeci and colleagues compared conventional endoscopy with chromoendoscopy in 109 patients using MB. The sensitivity of chromoendoscopy for Barrett epithelium was superior to that of conventional endoscopy (87% compared with 66%; P <.05): However, there was no statistical difference between the 2 methods in the diagnosis of esophageal carcinoma ( P >.05). They concluded that chromoendoscopy is useful for delineating Barrett epithelium and for indicating the correct location for securing biopsies when dysplasia or early esophageal cancer is suspected.

MB is a vital dye with intracellular binding characteristics, and safety issues have been raised regarding possible DNA damage associated with white-light illumination. These concerns, along with high cost, long procedure time, and unreliable detection of mucosal abnormalities, have impaired the widespread application of vital dye staining chromoendoscopy techniques in clinical practice.

Chromoendoscopy

Chromoendoscopy involves endoscopic examination of gastrointestinal mucosa after applying a dye solution to the surface, which enhances the appearance of otherwise nonperceivable mucosal changes. This technique guides the endoscopist to take target biopsies only from the suspicious areas. Different dye solutions are used including Lugol iodine solution, methylene blue (MB), indigo carmine, crystal violet, and acetic acid. Lugol solution consists of a 0.5% to 3.0% aqueous solution of potassium iodide, and iodine and is taken up by glycogen-containing cells. It is more useful for detecting dysplasia in patients with increased risk of squamous cell carcinoma (ie, with a history of smoking, alcoholism, and lye ingestion). On the other hand, MB, indigo carmine, and acetic acid are more useful in the detection of glandular abnormalities, as seen in adenocarcinoma. After staining with these agents, different mucosal patterns become visible during videoendoscopy. Various studies have been done to classify and detect dysplasia and early carcinoma based on these patterns. Guelrud and colleagues used acetic acid, a mucolytic agent, to stain the esophageal mucosa, and observed 4 pit patterns (round, reticular, villous, and ridged). Villous and ridged patterns were found to be associated with intestinal metaplasia. Sharma and colleagues studied the mucosal changes in 80 patients with Barrett esophagus (BE) after application of indigo carmine, and observed 3 mucosal patterns: ridged/villous, circular, and irregular/distorted. The irregular/distorted pattern was found to be associated with Barrett high-grade dysplasia (HGD) or superficial adenocarcinoma in 6 patients. In the same study, the presence of the ridged or villous pattern had 97% sensitivity, 76% specificity, and 92% positive predictive value (PPV) for prediction of intestinal metaplasia. However, in subsequent studies, these results could not be reproduced and no detection benefit for either Barrett metaplasia or dysplasia could be established with chromoendoscopy. Other studies using acetic acid and crystal violet stain also produced variable results. The probable reasons for these conflicting observations could be differences in technique, operator experience, and a patient population with a prevalence of BE. This was further substantiated in a blinded European study in which 4 expert gastrointestinal endoscopists analyzed magnification chromoendoscopy images of BE, using acetic acid or MB. The interobserver agreement was poor (κ = 0.40) for all parameters studied, including the mucosal patterns, MB positive staining, and the presence of specialized intestinal metaplasia (SIM).

Ngamruengphong and colleagues performed a meta-analysis of 9 studies published in PubMed for assessment of the diagnostic yield of techniques of chromoendoscopy compared with conventional 4-quadrant random biopsy (RB) in detection of SIM and dysplasia in patients with BE. A total of 450 patients with BE were reported in 9 studies included in the meta-analysis. Data on the yield of both modalities were extracted and analyzed to estimate weighted incremental yield (IY) and 95% confidence intervals (CIs) of MB chromoendoscopy with target biopsies over RB protocol using the Cochrane Q χ ² test. There was no significant IY with MB over RB for detection of SIM (IY 4%; 6 studies, n = 251), dysplasia (IY 9%; 9 studies, n = 450), and HGD or early cancer (EC) (IY 5%; 8 studies, n = 405). This meta-analysis shows that the technique of MB chromoendoscopy only has a comparable yield with white-light endoscopy using RBs for the detection of SIM and dysplasia during endoscopic evaluation of patients with BE.

Ormeci and colleagues compared conventional endoscopy with chromoendoscopy in 109 patients using MB. The sensitivity of chromoendoscopy for Barrett epithelium was superior to that of conventional endoscopy (87% compared with 66%; P <.05): However, there was no statistical difference between the 2 methods in the diagnosis of esophageal carcinoma ( P >.05). They concluded that chromoendoscopy is useful for delineating Barrett epithelium and for indicating the correct location for securing biopsies when dysplasia or early esophageal cancer is suspected.

MB is a vital dye with intracellular binding characteristics, and safety issues have been raised regarding possible DNA damage associated with white-light illumination. These concerns, along with high cost, long procedure time, and unreliable detection of mucosal abnormalities, have impaired the widespread application of vital dye staining chromoendoscopy techniques in clinical practice.

Narrow-band imaging

The working principal of narrow-band imaging (NBI) is based on the use of light filters which reduce red and green light and preserve the amount of blue light that illuminates the tissue. The resultant blue-green excitation light with narrow bandwidth has limited optical scattering and shallow penetration depth. It improves the imaging of mucosal and glandular changes and efficiently visualizes abnormal vascular patterns. As blue light is readily taken up by hemoglobin, NBI successfully detects the increased density of abnormal microvessels that is associated with dysplasia and carcinoma.

This technology was first introduced by Gono and colleagues in 1999 as a combined effort of the Japanese National Cancer Center Hospital East and a team of bio-optical physicists from Olympus Corporation (Tokyo, Japan). There are 2 versions of the NBI system. One version is called the Evis Exera II system, which is predominately used in North America. It is equipped with several diminutive band-pass color filters that allow green light of 530 to 550 nm bandwidth and blue light of 390 to 445 nm bandwidth to pass through and activate each pixel on a trichromatic charge-coupled device (CCD). The other system is the Lucera system, which uses a monochromatic CCD system and is used predominantly in Japan and Europe. Both systems allow switching between high-resolution white-light and NBI modes using a switch on the handle of endoscope. Although they have slight technical differences, the systems are functionally similar.

NBI has several advantages over vital dye chromoendoscopy in terms of applicability, cost, time, tidiness, and precision. It is widely available commercially and has already received regulatory approval. It is one of the most extensively studied advanced endoscopic imaging techniques for the detection of esophageal dysplasia and superficial carcinoma.

The mucosal and vascular patterns seen in BE have been the basis of many single-center studies ( Fig. 1 ). Kara and colleagues studied these patterns in BE patients and observed the association of regular mucosal and vascular patterns with intestinal metaplasia, whereas irregular mucosa and abnormal blood vessels were present in Barrett HGD. In another study, Kara and colleagues compared HRE with indigo carmine chromoendoscopy or NBI in 14 patients with Barrett HGD. The aim of the study was to compare and correlate the findings with the combinations of these techniques for the detection of Barrett HGD or superficial carcinoma. Eleven patients with HGD (79%) were detected with HRE alone, whereas NBI found HGD in 12 patients (86%). Indigo carmine was able to detect HGD in 13 patients (93%), whereas HGD (7%) was found in 1 patient using RB, which could not be detected with any imaging modality. NBI found an additional 4 HGD lesions in 3 of these 12 patients. White-light resolution endoscopy detected all cases of HGD, showing better efficacy for primary detection than NBI. However, NBI was more useful for detailed inspection of suspicious lesions. As a historical comparison, a previous study performed by this group detected HGD in 62% of patients using targeted standard resolution endoscopy (SRE) biopsies, and in 85% of patients with SRE-targeted plus random quadrantic biopsies.

A BE lesion with HGIN detected with AFI. ( A ) During inspection with white light, this area was not judged as suspicious. ( B ) The area around the small squamous island in the middle of the image showed a blue-violet autofluorescence imaging color. ( C, D ) With NBI, irregular and disrupted mucosal patterns were found. The histopathology confirmed the presence of HGIN. ( From Curvers WL, Singh R, Song LM, et al. Endoscopic tri-modal imaging for detection of early neoplasia in Barrett’s oesophagus: a multi-centre feasibility study using high-resolution endoscopy, autofluorescence imaging and narrow band imaging incorporated in one endoscopy system. Gut 2008;57(2):167–72; with permission.)

Anagnostopoulos studied mucosal and vascular features of Barrett disease in 50 patients with 344 lesions using high-resolution magnification endoscopy and NBI. The sensitivity, specificity, positive and negative predictive values of regular mucosal, and vascular patterns for intestinal metaplasia were 100%, 79%, 94%, and 100% respectively, whereas sensitivity, specificity, positive and negative predictive values for HGD in these patients were 90%, 100%, 99%, and 100%, respectively.

Curvers and colleagues performed a blinded study involving 14 patients with 22 suspicious lesions, including 8 lesions of HGD, 1 lesion with low-grade dysplasia, 1 lesion indefinite for dysplasia, and 12 areas of nondysplastic Barrett disease. They performed high-resolution endoscopy with vital dye staining techniques, using acetic acid and indigo carmine and NBI. In a blinded fashion, 7 community and 5 expert gastrointestinal endoscopists evaluated standard images from these lesions for any association of glandular and vascular patterns with dysplasia. The detection rate for dysplasia or neoplasia, with high-resolution white-light endoscopy, was 86% overall (90% for experts and 84% for nonexperts). The addition of vital dye staining or NBI did not improve the diagnostic yield. By contrast, Herrero and colleagues evaluated a simplified NBI classification of mucosal morphology to assess inter- and intraobserver agreement and the correlation with histology. Two hundred NBI images were evaluated twice by 4 endoscopists experienced in NBI and 4 inexperienced endoscopists at 2 referral centers. Endoscopists assessed each image for quality, suspicion for dysplasia, and regularity of mucosal and vascular patterns. Overall interobserver agreement was seen to be moderate (κ 0.42–0.44), whereas overall intraobserver agreement was moderate to substantial (κ 0.60–0.62). However, endoscopists in this study correctly identified 71% of the images containing HGD/EC and 68% of nondysplastic images were correctly identified as not suspicious. There were no significant differences in agreement between expert and inexpert endoscopists, suggesting a short “learning curve.” This low rate of identification of HGD/EC (71%) contrasts with the previous studies, cited previously, and Singh’s prospective single-center study involving 21 patients with BE, comparing the imaging characteristics between high-resolution magnification white-light endoscopy and NBI with histology. Mucosal patterns (pit pattern and microvascular morphology) were evaluated for their image quality on a visual analog scale of 1 to 10 by 5 expert endoscopists who then predicted mucosal changes based on pit and microvascular patterns. NBI was superior to white-light endoscopy in the prediction of histology in BE, as the overall pit and microvasculature quality was significantly higher ( P <.001). Furthermore, NBI was also observed to be superior to white-light endoscopy for prediction of dysplasia (χ ² = 10.3, P <.05). The overall κ agreement among the 5 endoscopists was 0.59 and 0.31 ( P <.001) showing good reproducibility. Although these “off-line” studies that review images from procedures have generally confirmed the usefulness of NBI for the diagnosis of Barrett disease and the detection of dysplasia, the low rate of identification of HGD/EC (71%) in the Herrero study raises questions regarding the ability of NBI to replace the use of RB protocols for surveillance endoscopy in BE patients.

In a prospective, blinded, tandem study, our group at Mayo Clinic compared SRE and HRE-NBI in 65 patients with BE. With HRE-NBI, dysplasia was detected in 37 patients (57%), whereas SRE with targeted plus RBs detected dysplasia in 28 patients (43%), which was statistically significant ( P <.001). NBI also found higher grades of dysplasia in 12 patients (18%), compared with no cases of SRE, with targeted plus RBs, detecting a high grade of histology (0%; P <.001). This study also showed greater efficiency of HRE-NBI, as fewer NBI-directed biopsies (mean 4.7 biopsies per case; P <.001) were needed to detect dysplasia in significantly more patients with BE, compared with SRE with targeted plus RBs (mean 8.5 biopsies per case). Sharma and colleagues presented the results of a prospective, multicenter, randomized crossover trial of HRE with quadrantic biopsy protocol compared with HRE with NBI-targeted biopsies for dysplasia detection in 116 patients undergoing screening or surveillance for BE. Overall, there was no significant difference between the 2 modalities in the primary aim of detecting intestinal metaplasia (HRE 85%, NBI 86%, P = .61). However, NBI detected HGD and cancer (23%) more often compared with HRE, even though the proportion of patients with neoplasia as not statistically different between the groups (HRE 29%, NBI 34%, P = .22). HRE detected 1 of 3 cancers and 7 of 10 HGD patients, whereas NBI detected all 3 cancers and 8 of 10 HGD patients. NBI also detected significantly more lesions with HGD/cancer than HRE (17 vs 10, P = .03) and more overall lesions with any degree of dysplasia compared with HRE (71 vs 55, P = .0002). NBI also required fewer biopsies per procedure (3.7 vs 8.0, P <.0001). This study showed that, although the overall rate of intestinal metaplasia detection was similar between HRE and NBI, NBI achieved this with fewer biopsies per procedure and had significantly better detection rates for neoplastic lesions. These clinical studies demonstrate the importance of HRE in combination with NBI for the surveillance evaluation of BE patients. Further, NBI is the most rigorously studied method of “virtual chromoendoscopy,” with controlled studies suggesting that the use of NBI improves the accuracy and efficiency of dysplasia detection in BE patients.