Interobserver agreement among endoscopists on evaluation of polypoid colorectal lesions visualized with the Pentax i-Scan technique

Abstract

Background and aims

Advances in colonoscopy, such as the Pentax i-Scan electronic technique, have the potential to improve the early detection of colorectal cancer. The aim of this multicentre study was to assess the interobserver agreement in the visualization of the surface and margins of colorectal polyps and in distinguishing neoplastic from non-neoplastic polyps.

Patients and methods

Eight expert endoscopists examined 400 mixed previously recorded images of polyps taken with different Pentax i-Scan settings in order to give an evaluation of the surface of the polyp and regular colonic mucosa, the pit-pattern and the nature of the lesion.

Results

A total of 400 mixed images of polyps with a diameter >5 mm and <10 mm were stored for analysis. Overall, there was a K_f agreement of 0.370 (p < 0.001) and 0.306 (p < 0.001) regarding pit-pattern and margins, respectively. The K_f agreement for the difference between neoplastic and non-neoplastic lesions was of 0.446 (p < 0.001).

Conclusions

We observed good interobserver agreement in the evaluation of neoplastic and non-neoplastic lesions and poor agreement in the evaluation of pit-pattern and margins. Adequate training is required in order to interpret images acquired with the i-Scan technique.

Introduction

Colorectal cancer (CRC) is the third cause of cancer death in the USA [1] and the fourth most common form of cancer in the world, as well as one of the most frequent in the Western world [2]. It is widely accepted that adenomatous polyps, which represent one half to two thirds of all colorectal polyps [3], [4], are precursors of CRC; multiple genetic alterations have been implicated in the evolution of adenoma to carcinoma [5]. Colonoscopy is regarded as the gold standard for diagnosing colorectal neoplasia [6], [7] and colonoscopy with polypectomy of neoplastic adenomatous lesions reduces the risk of CRC by as much as 80% for a period that may exceed 10 years [8]; furthermore, it has recently been reported that this strategy also prevents death from CRC [9]. Although colonoscopy is the gold standard screening modality for CRC, it does have some diagnostic limitations. Some studies have shown an overall neoplasm miss rate of up to 22% [10], [11].

Lesions suspicious of neoplasia and protruding adenomas are easy to identify with colonoscopy but the recognition of flat, depressed adenomatous lesions is sometimes difficult, especially when they are hidden by colonic folds, and is only possible with thorough, careful inspection [12]. Chromoendoscopy performed with magnifying endoscopes, although time-consuming, has been shown to identify colonic lesions easily, especially flat, small or depressed ones, and to predict histology reliably based on the surface pit-pattern architecture [13], [14].

New endoscopic techniques have been developed with the aim of improving the detection rate of small polyps and facilitating the differentiation between neoplastic and non-neoplastic polyps, with the possibility of visualizing the pit-pattern correctly. Advances in colonoscopy technology, such as Fuji intelligent chromoendoscopy (FICE), narrow band imaging (NBI) and the Pentax i-Scan electronic technique have the potential to improve the early detection of CRC. The i-Scan technique is a new, high-definition (HD+) image-enhanced endoscopy technique, with an electronic post-processing light filter using different software algorithms and real-time image mapping (RIM).

Although electronic chromoendoscopy techniques such as NBI and FICE are useful for detecting and diagnosing early neoplastic lesions with or without magnifying endoscopy [15], these techniques yield images that are much darker than conventional white-light images, particularly in regions of the gastrointestinal tract with a large lumen. Pentax i-Scan images are as bright as conventional, white-light images and, compared to FICE and NBI, are able to visualize much larger areas in a distant view and, unlike NBI, do not require staying close to the bowel wall. Furthermore, this Pentax technique does not need endoscopic magnification for the observation of the margins of lesions. There are, however, few studies in the literature about this new technique and, in particular, about which endoscopic Pentax i-Scan settings are most helpful in the differentiation between neoplastic and non-neoplastic colonic lesions.

Pentax i-Scan HD+ is a new, fully digital, high-definition, video-endoscopic system. The particular characteristics of the new Pentax Hi Line HD+ endoscopes, compared to those of the previous series, are that a megapixel CCD chip contains about 1,300,000 pixels compared to the approximately 410,000 pixels in the standard, digital zoom, which can double the magnification of the image. The EPK-i video processor processes a high-definition video (DVI-D) signal output with a progressive scan image that simultaneously moves all the lines and the pixels of the image. The system can achieve an endoscopic image resolution about three times higher than that of a standard endoscope. The video processor also offers the functions of the i-Scan, a filter system that can enhance the scanned images in real-time by modifying the software.

The i-Scan technology combines brightness of light (on a scale from −5 to +5) with noise reduction (low, medium and high) and three types of enhancement of the scanned image (surface enhancement – SE; contrast enhancement – CE; tone enhancement – TE) which, thanks to real-time changes in software algorithms, can be used individually or in combination. These algorithms, by emphasizing the virtual colour and contrast of the mucosa, provide a better view of the vascular patterns and architecture of the mucosa thus improving the identification of subtle surface irregularities of the mucosa and vasculature. These algorithms could, therefore, facilitate the identification of pre-cancerous lesions, even small ones. The more uniform lighting system and an image projected on a full screen give a new, improved depth of endoscopic view of the field.

With SE, the differences in luminance intensity (scale: from 4 to 6) between the pixels concerned and the surrounding pixels is analysed and the edge components are enhanced. The adjustment of the noise erasure function provides greater enhancement of the edges, corresponding to small variations in structure. Moreover, SE improves the visualization of minute glandular structures, making it easier to identify changes on the basis of structural differences.

With CE, areas of lower luminance (scale: from 2 to 6) are identified on the basis of pixel-wise luminance intensity data, followed by relative enhancement of the blue (B) component through slight suppression of the red (R) and green (G) components. The low luminance areas are then coloured slightly bluish-white and minute irregularities on the mucosal surface are enhanced. Irregular glandular structures in flat mucosa are also enhanced.

This function electronically substitutes the vital staining with various dyes used in endoscopy. The white-light images are processed and broken down into three main colour components (red, green and blue). The EPK-i works on the dynamics of the electronic components of colour (primarily the red component) through an algorithm and processes the image by enhancing the desired colour. TE functions available are similar to those of chromoendoscopy and are divided into TE p (pit-pattern), v (vessels), e (oesophagus), b (Barrett), g (gastric) and c (colonic).

Until now, no studies on interobserver agreement using the i-Scan technique have been reported. The aim of this work was to assess the interobserver agreement on the visualization of the surface (identifying the colonic pit-pattern) and margins of polyps and in distinguishing neoplastic from non-neoplastic colorectal polyps, based on the interpretation of the pit-pattern.