Roel Bogie

Development and validation of an educational web-based system for endoscopic classification of laterally spreading tumors

2

Figure 2.2: Illustration of the endoscopic Kudo classification. | A: LST-G-H, B: LST-G-NM, C: LST-NG-FE, D: LST-NG-PD. Licensed by Erik Wallert (www.erikwallert.nl).

cases with anonymized clinical data and endoscopic images were kindly provided by two of the study investigators (HMC, SS). High-definition white-light colonoscopy and chromoendoscopy images were available in all cases. Both Olympus (H260 or H290 series) and Pentax (EC-3890i series) colonoscopes were used. To ensure uniformity in classification of images from the 2 centres, we selected only endoscopic images using dye-based chromoendoscopy (indigo carmine 0.04%). Endoscopic images using digital chromoendoscopy techniques were not used. For each case photo documentation comprised long-view images to assess the location and size of the LST and close view images to assess the endoscopic shape (at least 2 per case). All images were taken in a clean and well-insufflated colon before starting the resection. Post-resection histopathology assessed by experienced gastroenterology pathologists was available but not presented to raters. Sample size calculation indicated that minimal 50 cases were needed for a 95% confidence interval of +-0.10 in case of a suspected kappa coefficient of 0.75. 15 We invited colonoscopy experts from Western and Eastern countries and colonoscopy trainees from MUMC+ and NTUH in the study. We included at least 6 raters in each group. The precision of the kappa coefficient does not increase much more when including more than 6 raters. 15 We assumed an equal distribution of the endoscopic Kudo subtypes of LSTs. To facilitate sensitivity analysis, a total number of 72 cases were included. From the 103 cases selected by the study investigators out of their personal selection, eventually 72 were selected to use in the modules. Selection was based on a quality-assessment comprising clarity of the image, the presence of air bubbles, presence of long-view and short-view, level of air insufflation and use of blue dye. As in daily practice, multiple cases had flaws on some of the points. Therefore, the core-team selected the 72 cases with the highest overall quality. Fourteen international experts in colonoscopy were invited to participate (Step 2). All experts perform research in image-enhanced endoscopy-assisted diagnosis and endoscopic resection of

21