Background and Aims. Linked color imaging (LCI) helps screen and diagnose for early gastric cancer by color contrast in different mucosa. RGB (red, green, and blue) pixel brightness quantifies colors, which is relatively objective. Limited studies have combined LCI images with RGB to help screen for early gastric cancer (EGC). We aimed to evaluate the RGB pixel brightness characteristics of EGC and noncancer areas in LCI images. Methods. We retrospectively reviewed early gastric cancer (EGC) patients and LCI images. All pictures were evaluated by at least two endoscopic physicians. RGB pixel brightness analysis of LCI images was performed in MATLAB software to compare the cancer with noncancer areas. Receiver operating characteristic (ROC) curve was analyzed for sensitivity, specificity, cut-off, and area under the curve (AUC). Results. Overall, 38 early gastric cancer patients were enrolled with 38 LCI images. Pixel brightness of red, green, and blue in cancer was remarkably higher than those in noncancer areas (190.24±37.10 vs. 160.00±40.35, p<0.001; 117.96±33.91 vs. 105.33±30.01, p=0.039; 114.36±34.88 vs. 90.93±30.14, p<0.001, respectively). Helicobacter plyori (Hp) infection was not relevant to RGB distribution of EGC. Whether the score of Kyoto Classification of Gastritis (KCG) is ≥4 or p>0.05). Receiver operating characteristic (ROC) curve for differentiating cancer from noncancer was calculated. The maximum area under the curve (AUC) was 0.767 in B/G, with a sensitivity of 0.605, a specificity of 0.921, and a cut-off of 0.97. Conclusions. RGB pixel brightness was useful and more objective in distinguishing early gastric cancer for LCI images.

The early detection accuracy of early gastric cancer (EGC) determines the choice of the optimal treatment strategy and the related medical expenses. We aimed to develop a simple, affordable, and time-saving diagnostic model using six machine learning (ML) algorithms for the diagnosis of EGC. It is based on the endoscopy-based Kyoto classification score obtained after the completion of endoscopy and other clinical features obtained after medical consultation. We retrospectively evaluated 1999 patients who underwent gastrointestinal endoscopy at the China Beijing Hospital. Of these, 203 subjects were diagnosed with EGC. The data were randomly divided into training and test sets (ratio 4:1). We constructed six ML models, and the developed models were evaluated on the testing set. This procedure was repeated five times. The Kolmogorov–Arnold Networks (KANs) model achieved the best performance (mean AUC value: 0.76; mean balanced accuracy: 70.96%; mean precision: 58.91%; mean recall: 70.96%; mean false positive rate: 26.11%; mean false negative rate: 31.96%; and mean F1 score value: 58.46). The endoscopy-based Kyoto classification score was the most important feature with the highest feature importance score. The results suggest that the KAN model, the optimal ML model in this study, has the potential to identify EGC patients, which may result in a reduction in both the time cost and medical expenses in clinical practice.