Texture and Colour Enhancement Imaging versus White Light Endoscopy for Detection of Dysplasia within Barrett’s Oesophagus: A Pilot Study Open Access

Oesophageal cancer is a significant global health concern, ranking as the seventh most common cancer and the sixth leading cause of cancer-related mortality worldwide [1]. In 2020 alone, there were an estimated 604,000 new cases and 544,000 deaths attributed to this malignancy, reflecting its poor prognosis, particularly when diagnosed at a late stage [1]. Despite advances in screening, prevention, diagnosis and treatment, the 5-year survival in developed countries has only risen from less than 5% in the 1960s to 20% over the last decade [2]. While the global incidence of oesophageal squamous cell carcinoma has decreased, there has been a marked increase in oesophageal adenocarcinoma incidence in Europe, North America, and Australia [3]. This has been accompanied by a rise in the prevalence of gastroesophageal reflux disease, obesity and accordingly Barrett’s oesophagus (BE), a precursor lesion to adenocarcinoma [4‒6].

BE affects more than 3% of the population and progresses in a stepwise fashion from non-dysplastic BE, to low-grade dysplasia (LGD), to high-grade dysplasia (HGD) and eventually adenocarcinoma [7‒9]. The likelihood of progression to adenocarcinoma increases as dysplasia advances, with an incidence rate of 6.58 cancers per 100 patient-years in those with previous HGD [7]. This stepwise development presents an opportunity for detection and intervention as dysplasia progresses prior to the development of more advanced disease. Despite this, a 2016 meta-analysis found that in BE surveillance cohorts with initial non-dysplastic BE or LGD, approximately 25% of adenocarcinomas are diagnosed within 1 year of the initial endoscopy and therefore are considered missed [10]. This has led to the development of surveillance biopsy protocols, with the most universally utilized being the Seattle protocol, involving 4-quadrant biopsies every 2 cm from the involved BE segment (or 1 cm in the context of known dysplasia) [11]. Nevertheless, such protocols are costly and time-consuming and still risk missing up to 30% of HGD when using white light endoscopy (WLE) [12].

Multiple forms of advanced mucosal imaging have been explored as potential solutions to this dilemma, with narrow-band imaging (NBI) the most widely researched, detecting more dysplasia with fewer biopsies than WLE and Seattle protocol biopsies [12]. However, while similar data does not yet exist for BE, studies in endoscopy and colonoscopy have highlighted the significant learning curve for accurate application of NBI which renders its usefulness in many cases exclusive to the expert interventional endoscopist [13, 14]. This may relate in part to the considerable change in colour spectrum when using NBI compared to WLE, the most commonly used imaging modality in routine endoscopy (Fig. 1). Texture and Colour Enhancement Imaging (TXI) is a new push-button imaging technology from Olympus that aims to improve lesion delineation without significantly altering the colour spectrum of the WLE image (Fig. 2). During processing, the incoming image is split, with brightness enhanced in the base layer image and texture enhanced in the detail layer, before the images are merged. In TXI mode 1 the final image undergoes further colour enhancement while in mode 2 it does not (Fig. 3) [15]. The aim of this study was to determine whether TXI could be a valuable addition to the armoury of advanced mucosal imaging techniques for BE surveillance.

This pilot study was performed at a single South Australian tertiary centre from 2021 to 2024. All procedures were performed by a single experienced interventional endoscopist. Procedural sedation was used except where patients were deemed to require a general anaesthetic by the treating anaesthetist. The Olympus EVIS X1 endoscopy system using GIF-HQ190 and GIF-EZ1500 video-gastroscopes (Olympus Medical Systems Co., Tokyo, Japan) were used in all cases. All patients aged 18–80 years undergoing surveillance for BE by a single interventional endoscopist were suitable for inclusion in the study, provided the length of the BE segment was at least 1 cm. For the purpose of this study, BE was defined according to the 2011 American Gastroenterological Association Position Statement as columnar epithelium with histologically proven intestinal metaplasia [16]. Patients were required to be on acid-suppressive therapy with a proton-pump inhibitor (PPI) at a minimum dose of pantoprazole 40 mg daily or equivalent, for 4 weeks prior to endoscopy. Patients who were unable to consent, had coagulation disorders or were taking anticoagulation, had significant comorbidities (severe heart failure, chronic kidney disease, or chronic obstructive pulmonary disease), or were pregnant were excluded from the study.

This was a randomized cross-over study (Fig. 4). Patients were randomized 1:1 to initial evaluation with either WLE or TXI mode 1. Randomization was performed by a random number generator with the assigned imaging mode for each study number concealed in a sealed envelope until after the patient had consented to participation in the study. After consent to the procedure and the study, patients were assigned consecutive study numbers which then correlated to their assigned imaging mode. After careful evaluation of the entire BE segment on the selected initial imaging modality, any areas of concern were documented on a case record form according to distance from the incisors and position on a clock-face. “Areas of concern” included nodularity or depression, as well as alterations in vascularity or pit pattern (examples in Fig. 5). An additional evaluation was then performed with the unused imaging mode (WLE or TXI mode 1) and again any areas of concern documented. Finally, NBI with magnification (M-NBI) was used to further assess the entire BE segment, though focusing primarily on identified areas of concern. Targeted biopsies were then taken of all areas of concern seen on any imaging modality. Routine surveillance biopsies were then taken in a 4-quadrant fashion at 1-cm intervals throughout the entire BE segment. Patients were given oral simethicone just prior to the procedure to aid visualization, and all procedures were performed using a transparent cap attachment.

Data collected included demographic data, PPI usage and dose, the duration and length of BE segment (according to the Prague Classification), the number of areas of concern noted endoscopically and the imaging modalities on which these areas were visualized, and finally the histology for all targeted and random biopsies [17]. All pathology specimens were reviewed by two gastrointestinal pathologists, with the final consensus diagnosis adjudicated by a local expert BE histopathologist. As this was a pilot study, descriptive statistics were reported, with limited numbers preventing further statistical analysis. This study was prospectively registered with the Australian New Zealand Clinical Trials Registry (ANZCTRN: 12,621,000,893,808).

A total of 52 patients were recruited for this pilot study. Two patients were excluded due to a length of BE less than 1 cm, leaving 50 patients for the final analysis. Ninety-two percent of patients (n = 46) were male with a median age of 68.5 years. The median length of BE according to the Prague Classification was C2M3 with a median disease duration of 3 years. A total of 56% procedures (n = 28) were routine BE surveillance, while 16% (n = 8) were follow-up procedures after endoscopic resection, 20% (n = 10) were patients referred with LGD on random biopsies and 8% (n = 4) were referred with HGD on random biopsies. All patients were on standard-dose PPI at the time of procedure, with 7 patients still having active inflammation at the time of endoscopy.

There were 27 suspicious lesions seen endoscopically in 22 (44%) patients, with endoscopic lesion characteristics as described in Table 1. On targeted biopsies, 25.9% (n = 7) were early adenocarcinoma, 25.9% (n = 7) were HGD, 18.5% (n = 5) were LGD, 18.5% (n = 5) were inflammatory and 11.1% (n = 3) were uninflamed non-dysplastic BE. Seattle protocol biopsies were then performed for all patients and identified HGD in 1 patient and LGD in 14 patients. 73.3% (n = 11/15) of HGD and early adenocarcinoma were visible on WLE (Table 2). A total of 93.3% (n = 14/15) HGD or early adenocarcinoma were visible on assessment with both TXI and NBI, with 1 patient having flat multifocal HGD on Seattle protocol biopsies. Importantly, this patient had a concurrent area of HGD that was endoscopically visible on assessment with TXI and NBI, and therefore per-patient sensitivity and NPV remained 100%. In 1 patient with an 8 mm early adenocarcinoma and 2 patients with HGD (6 mm and 12 mm), a lesion of concern was seen on TXI that had not been detected using WLE. In all three cases, these lesions were characterized by subtle nodularity and altered vascularity which was made more pronounced by TXI, thereby defining the lesions more clearly from the surrounding BE mucosa (Fig. 6). A total of 28% (n = 14/50) patients had LGD detected on Seattle protocol biopsies with no identifiable lesions. Overall, the detection of LGD using any imaging format was poor, with only 15.8% (n = 3/19) of LGD detected on WLE and 26.3% (n = 5/19) detected on TXI or NBI.

For the detection of HGD/early adenocarcinoma, TXI had a sensitivity of 93.3% (95% CI, 68.1–99.8%), specificity of 69.2% (95% CI, 52.4–83%), NPV of 96.4% (95% CI, 80.2–99.5%), and accuracy of 75.9% (95% CI, 62.4–86.5%) (Table 3). Comparatively, WLE had a sensitivity of 73.3% (95% CI, 44.9–92.2%), specificity of 75% (95% CI, 56.6–88.5%), NPV of 85.7% (95% CI, 71.7–93.4%), and accuracy of 74.5 (95% CI, 59.7–86.1%). Histological prediction after further assessment with M-NBI had a sensitivity of 93.3% (95% CI, 68.1–99.8%), specificity of 81.8% (95% CI, 64.5–93%), NPV of 96.4% (95% CI, 80.2–99.5%), and accuracy of 85.4% (95% CI, 72.2–93.9%). On a per-patient analysis, TXI alone had a sensitivity of 100% (95% CI, 71.5–100%), specificity of 71.8% (95% CI, 55.1–85%), NPV of 100% (95% CI, 87.7%–100%), and accuracy of 78% (64–88.5%); however, the combination of TXI and with M-NBI improved the specificity to 84.6% (95% CI, 69.5–94.1%) and the accuracy to 88% (95% CI, 75.7–95.5%) (Table 4).

This study has demonstrated the feasibility of using TXI for detection of dysplasia in BE. While statistical analysis is limited due to the sample size, there appears to be a percentage of subtle BE dysplasia and even early adenocarcinoma that is not readily visualized on WLE but can be detected when using TXI and/or NBI. For interventional endoscopists, this may improve detection of previously “invisible” dysplasia to guide appropriate treatment strategies. For routine endoscopists and in the community setting, this may improve the yield of targeted and Seattle protocol biopsies.

As is demonstrated in this study, with appropriate equipment and expertise more than 90% of HGD/early adenocarcinoma in BE is visible endoscopically using TXI and/or NBI. This is similar (though numerically superior) to the pooled sensitivity of 83% (95% CI, 73–93%) in a 2021 meta-analysis using NBI [18]. This high degree of sensitivity is critical when considering endoscopic therapy in BE. Endoscopic eradication therapy is indicated for flat LGD or HGD in BE in the absence of endoscopically visible lesions, with radiofrequency ablation (RFA) remaining first-line [19]. However, RFA is inadequate treatment for adenocarcinoma and would not only result in non-curative therapy but would risk “burying” of the malignant tissue beneath neo-squamous epithelium. Even in the presence of missed HGD, treatment with RFA has been associated with buried dysplastic BE glands which can progress and remain obscured by overlying neo-squamous epithelium [20, 21]. Accordingly, prior to considering any endoscopic intervention, proceduralists should thoroughly assess the entire BE segment for visible lesions, with TXI being a useful addition to the armamentarium in this setting.

In the routine endoscopy setting, TXI may improve the accuracy of biopsies by non-interventional endoscopists, without significantly altering the familiar colour spectrum of WLE. While studies have demonstrated that dedicated Barrett’s surveillance endoscopists achieve higher dysplasia detection rates than community endoscopists, in many regions resource constraints limit patients’ access to dedicated interventional endoscopists [22]. Therefore, training and technology that may serve to optimize dysplasia detection for routine endoscopists may lead to a reduction in missed oesophageal cancers. While NBI has been conclusively demonstrated to improve dysplasia detection, the colour spectrum is significantly altered from WLE and studies in gastric and colonic lesions have highlighted the learning curve required for effective use [12‒14, 18]. By maintaining a familiar colour spectrum to WLE, the learning curve for TXI may be comparatively short, as has now been demonstrated in adenoma detection during colonoscopy [23]. The potential benefit of the use of TXI in community BE screening is an area in which further research is required.

Additionally, this study again reinforces that LGD remains poorly delineated endoscopically despite advances in mucosal imaging technology. This is of importance as according to guidelines, patients with LGD require early follow-up and eventually endoscopic therapy for eradication if the LGD persists [24]. However, the clinical relevance of endoscopically invisible LGD has become increasingly contentious. A 2011 study by Wani et al. [25] demonstrated that the risk of progression to early adenocarcinoma was 0.44% per year and HGD 1.83% per year with no significant increase compared to non-dysplastic BE. There is significant interobserver variability in LGD diagnosis even among expert pathologists, leading to extremely heterogenous data on progression rates [26]. In addition, a high proportion of patients referred to expert centres with previously diagnosed LGD have more advanced dysplasia at endoscopy, hence data on progression rates is likely overestimated due to missed advanced dysplasia [27]. Accordingly, the American Society for Gastrointestinal Endoscopy has defined PIVI thresholds for the required accuracy of advanced imaging techniques if Seattle biopsy protocols are to be avoided [28]. Importantly, in our per-patient analysis (Table 3) the combination of TXI with subsequent characterization on M-NBI surpassed the PIVI thresholds of sensitivity >90%, NPV >98% and specificity >80% [28]. With further studies these advanced mucosal imaging modalities may therefore eventually obviate the need for non-targeted protocolized biopsies in BE.

This pilot study has highlighted the difficulties in performing randomized controlled trials in BE given the rarity of advanced dysplasia in this context. The limited numbers prevent any statistical conclusions being drawn from this analysis. In order to assess the potential benefit of TXI for BE surveillance and dysplasia detection with definite statistical conclusions, large multi-centre collaborative studies would be required.

In summary, this pilot study highlights the feasibility of TXI as an addition to the armamentarium for endoscopists performing surveillance and endoscopic therapy for patients with BE, demonstrating its ability to detect a portion of subtle lesions that had remained undetectable after assessment using WLE. Additionally, we have reinforced the high degree of sensitivity for HGD and early adenocarcinoma that can be achieved with the appropriate equipment and expertise in order to ensure the selection of appropriate endoscopic therapies. Large multi-centre collaborative studies will be required to conclusively demonstrate the effect size of TXI in this context.

This study was reviewed and approved by the Central Adelaide Local Health Network (CALHN) Human Research Ethics Committee (reference number 14671). Written informed consent was obtained from participants to participate in the study.

No authors have any conflicts of interest to declare.

Equipment was provided by Olympus Australia Ptd Ltd for the purposes of this study; however Olympus did not have any role in the study design or conduct.

All authors made meaningful contributions to this manuscript. E.Y. was responsible for recruitment, data collection, data analysis, manuscript writing, and submission. H.P. and R.S. were responsible for recruitment, manuscript review, and approval of the final manuscript for submission.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author (E.Y.) upon reasonable request.