OBJECTIVES: Computer-aided detection systems (CAD EYE EW10-EG01; Fujifilm, Tokyo, Japan) approved for clinical use in Japan have been developed to support endoscopists in the detection of gastric neoplasms during esophagogastroduodenoscopy (EGD). This prospective study aimed to evaluate the clinical performance of CAD EYE for gastric neoplasms by using white-light imaging (WLI) and linked color imaging (LCI) under real-world clinical conditions. METHODS: This multicenter, non-randomized, confirmatory trial enrolled patients with histologically confirmed gastric neoplasms at four Japanese institutions. Each lesion was sequentially observed during EGD using WLI-CAD EYE and LCI-CAD EYE. The primary endpoint was the sensitivity of the CAD EYE response for known gastric neoplasms, and the secondary endpoints included the characteristics of lesions without CAD EYE response and newly identified lesions. RESULTS: A total of 174 lesions were analyzed using the WLI-CAD EYE and 173 using the LCI-CAD EYE. The overall sensitivities were 92.0% (95% confidence interval CI, 86.9%-95.5%) and 96.5% (95% CI, 92.6%-98.7%), respectively. The lower bound of the 95% CI for LCI-CAD EYE exceeded the prespecified threshold of 90%, fulfilling the confirmatory criteria. WLI-CAD EYE and LCI-CAD EYE responses were observed in 57.4% and 59.2% of known lesions in a distant view, respectively. WLI-CAD EYE showed relatively low sensitivity for lesions < 10 mm (88.9%) and those located on the anterior wall (79.4%), whereas LCI-CAD EYE achieved ≥ 90% sensitivity in all subgroups. CONCLUSIONS: LCI-CAD EYE demonstrated a high sensitivity for known gastric epithelial neoplasms in a clinical setting.
Dohi et al. (2026) studied this question.