br Diagnostic tests br The diagnostic test
The diagnostic test consisted of 40 gastric lesions (22 cancerous and 18 noncancerous lesions). The characteristics of the 40 lesions are shown in Table 1. Three diagnostic tests (tests 1, 2, and 3) were composed of the same 40 gastric lesions. All M-NBI images of the lesions were taken at maximum magnification. Pathologic diagnosis of specimens was determined through endoscopic submucosal dissection in 22 cancer lesions, and 18 cases of noncancerous lesions were pathologically diagnosed by biopsy specimens. Conventional white-light images were not included in the diagnostic tests and were masked from participants. Endoscopic images of 40 lesions used for diagnostic tests were taken by an expert endoscopist (H.D.) at Ishikawa Prefectural Central Hospital. All M-NBI images were taken by magnifying endoscopy (GIF-Q240Z and GIF-H260Z; Olympus Co, Tokyo, Japan) fitted with a soft black hood. A light source was used (CV-260SL; Olympus Co), and the structure’s enhance setting during NBI observation was set to B8 level by fixing the NBI color mode to level 1.
Lesions recorded with image quality adequate for diag-nosis were selected for diagnostic tests. No biopsy sam-pling had been performed before the M-NBI examination in all 40 lesions.
Each participant evaluated the M-NBI images according to the VSCS. First, participants checked for the presence of the demarcation line (DL) and then evaluated the micro-vascular (MV) and microsurface (MS) patterns. The MV and MS patterns were classified into regular, irregular, and absent. Finally, each participant made an M-NBI diag-nosis for all lesions and judged these Talaporfin sodium (ME2906) lesions as cancerous or noncancerous according to the VSCS. Figure 2 shows examples of the representative M-NBI images used in the tests. Participants looked at images on the web, evaluated the DL and the MV and MS patterns, and made a diagnosis of cancer or noncancer according to the VSCS.
The e-learning system The e-learning system comprises 3 sections8:
1. The first video lecture discusses the principles of M-NBI and the basics of the VSCS. The video is approximately 40 minutes in duration.
2. The second video lecture discusses the practical applica-tions of the VSCS and the latest M-NBI findings. The video is approximately 90 minutes in duration.
3. The self-exercise tests comprise 100 M-NBI images (50 cancerous and 50 noncancerous lesions), with only 1 image for each lesion. There is Consensus sequemce no overlap of the images of the self-exercise tests and diagnostic test.
2 GASTROINTESTINAL ENDOSCOPY Volume -, No. - : 2019 www.giejournal.org
Ikehara et al Learning effect after e-learning training in endoscopic diagnosis
Figure 1. Study flow.
Area under the receiver operating characteristic curve of cancer diagnostic rate using M-NBI and interobserver agreement in the study participants
Between tests 1 and 2, about half of the participants learned by e-learning. Between tests 2 and 3, the other half learned by e-learning. Therefore, each area under the receiver operating characteristic curve (AUC/ROC) and the interobserver agreement of the answers of 365 par-ticipants were compared between baseline tests 1 and 3, in which all participants completed e-learning. The character-istics of the 40 lesions were classified according to macro-scopic type (elevated, flat, and depressed) and lesion size
(<10 and 10 mm). Each AUC/ROC was compared between tests 1 and 3. The interobserver agreement of the answers of the 365 participants was assessed, and the results between tests 1 and 3 were compared.
Continuous variables were tested for normal distribu-tion by the D’Agostino skewness test. The comparison of the changes in accuracy between tests 1 and 3 was calcu-lated using a paired t test. The diagnostic performance of M-NBI was evaluated by calculating the AUC/ROC, and the improvement in predictive accuracy was evaluated by calculating the continuous net reclassification
www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 3
Learning effect after e-learning training in endoscopic diagnosis Ikehara et al
TABLE 1. Characteristics of the 40 gastric lesions used for diagnostic tests
Well-differentiated adenocarcinoma 17 (77.3)
Moderately differentiated 5 (22.7)
Depth of invasion
Values are n (%) unless otherwise defined.
NA, Not applicable; SD, standard deviation.
improvement (cNRI) and integrated discrimination improvement (IDI), as described by Pencina et al.11 Each test was 2-sided, and P < .05 was considered statistically significant. Interobserver agreement was evaluated by calculating Fleiss’ kappa value.12 Kappa values of .01 to