Background and Objectives: Narrow band image (NBI) has already been known as a qualified endoscopic technique to improve detection of diverse upper aerodigestive tract lesions. But, relatively high cost and narrow applicability have made NBI could not...
Background and Objectives: Narrow band image (NBI) has already been known as a qualified endoscopic technique to improve detection of diverse upper aerodigestive tract lesions. But, relatively high cost and narrow applicability have made NBI could not be widely used. So, as an alternative way, we developed an image processing algorithm of RGB channel substitution which simulated the mechanism of NBI. The objective of this study is to present the algorithm of RGB channel substitution and examine the clinical applicability.
Materials and Methods: By the algorithm of the RGB channel substitution, RGB value in each pixel of endoscopic white light image is substituted; red value to green one, green value to blue one. This substitution realized an image similar with an image made by NBI filter. To examine clinical applicability of this algorithm, WLI and NBI of 50 benign (Barret’s esophagus, esophagitis) and 50 malignant (squamous cell carcinoma) esophageal lesions were selected, and WLI were converted to new images (CI). To eliminate systemic bias, the order of three groups of images (WLI, CI, NBI) was block-randomized. Three gastroenterologists independently evaluated randomized images without any clinical information. The accuracy, positive and negative predictive value, sensitivity and specificity of WLI, CI and NBI were analyzed.
Results: The accuracy was as following; WLI 84.3%, CI 86.3%, NBI 86.6%. The positive predictive value was as following; WLI 86%, CI 88.7%, NBI 85.3%. The negative predictive value was as following; WLI 82.8%, CI 84.3%, NBI 88.2%. The sensitivity was as following; WLI 82%, CI 83.3%, NBI 88.6%. And the specificity was as following; WLI 86.6%, CI 89.3%, NBI 84.6%. As a comparison of sensitivity, p-value was as following; WLI-CI 0.815, WLI-NBI 0.052, NBI-CI 0.134. As a comparison of specificity, p-value was as following; WLI-CI 0.344, WLI-NBI 0.581, NBI-CI 0.065. Despite no statistical difference of sensitivity and specificity, there was some trend of increasing accuracy in the order of WLI, CI and NBI. Ten of 150 examinations for malignant cases (6.6%) were correctly diagnosed by CI while their diagnoses by WLI were false-negative.
Conclusion: By substitution of RGB channels, WLI could be successfully converted to new images resembling NBI images. Although statistical significance could not be obtained in this study, RGB substitution could provide additional information on original WLI. An application program of RGB substitution can be applied to any kind of medical imaging tools (digital camera, microscopy, etc.) including endoscopy.
Key words: endoscopy, NBI, RGB channels