Introduction Narrow band imaging (NBI) utilises optical filters to enhance the visibility of fine mucosal microvasculature. Wavelengths of light used in NBI are shorter (between 415–540 nm) corresponding to the wavelengths most intensely absorbed by haemoglobin. We aimed to quantitatively compare the effect of improvement in contrast enhancement using NBI among various series of Olympus endoscopes (240, 260 and 290 series) using serial dilutions of haemoglobin.

Method We compared the visibility of human blood diluted with increasing concentrations (from ½ to 1/2²³) of distilled water in a 24 well transparent plastic plate. Still images of the wells (overview and individual) were captured using GIF Q240 and GIF H260 endoscopes with a LUCERA spectrum CV260 processor (Olympus, Keymed, UK) and GIF H290 endoscope with an ELITE CV 290 processor (Olympus, Keymed UK) with white light and NBI. Images were independently analysed by three groups of endoscopists for the presence of blood; endoscopists with experience in NBI use (NBI-experienced), endoscopists with no former NBI experience but who have obtained accreditation by Joint Advisory Group (JAG) for diagnostic endoscopies (NBI-Naive) and doctors with no prior endoscopy experience (novices). Inter-rater agreement was assessed using Kappa statistics. The median visible dilution was defined as the limit of visibility for the presence of haemoglobin in the overview image.

Results 45 participants (15 each of NBI-experienced, NBI-naïve and novices) completed the study. With conventional white light the median dilution at which all 3 groups noted the presence of haemoglobin was 1/2¹⁵ using all three generations of endoscopes. This improved using NBI to 1/2¹⁷for GIF Q240, 1/2¹⁹for GIF H260 and 1/2 ²⁰for GIF H290 series of endoscopes. The kappa values for inter-rater agreement was substantial for all 3 generations of endoscopes using conventional white light (κ >0.7) while it was only moderate to fair for all 3 generations of NBI (κ >0.5).

Conclusion NBI more effectively enhanced the contrast of diluted blood compared to conventional white light. These differences improved with successive generations of NBI but not with conventional white light. Inter-rater agreement was fair for NBI, but substantial for conventional white light. Effective training and more objective criteria for the presence of blood could help improve inter-observer agreement. Studies are needed to assess the clinical utility of newer generation NBI in the detection of early neoplasia and assessment of inflammation.

Disclosure of interest None Declared.