Objective To construct an artificial intelligence diagnostic system based on multi-feature fitting for diagnosing gastric whitish neoplastic lesions under white-light endoscopy. Methods Gastroscopic images from Renmin Hospital of Wuhan University and The Seventh Medical Center of Chinese PLA General Hospital were collected between December 2012 and July 2021. A total of 823 images of gastric whitish lesions from 267 patients were finally selected based on the inclusion and exclusion criteria. Five white-light endoscopic features that are associated with gastric whitish lesions were selected through a literature search: the location of the lesion (upper-middle stomach \ lower stomach \ undistinguished), the boundary of the lesion (clear \ unclear), the surface of the lesion (rough \ smooth), the roundness of the lesion (nearly circular \ non-circular), whether the lesion was depressed (depressed\ non-depressed). The models were trained by feeding images with manually labelled features into a machine learning algorithm for fitting, and selecting the optimal model as our multi-feature fitting diagnostic system; A conventional single deep learning model was trained with the same dataset. The diagnostic performance of the two models were compared, and six endoscopists at different levels were invited to make a human-computer comparison. Results The accuracy, sensitivity, and specificity of the multi-feature fitting diagnostic system were 82.11%, 78.43%, and 84.72%, respectively. The lesion characteristics in descending order of weighting were whether the lesion was depressed (weighting 0.71), the location of the lesion (weighting 0.11), whether the surface of the lesion was rough (weighting 0.08), whether the boundary of the lesion was clear (weighting 0.06), and whether the lesion was subcircular (weighting 0.04). The diagnostic accuracy of the multi-feature fitting diagnostic system was significantly higher than that of non-expert endoscopists (82.11% vs 74.31%, p= 0.008) and comparable to that of expert endoscopists (82.11% vs 83.20%, p= 0.700). There was no significant difference in accuracy between the multi-feature fitting diagnostic system and the traditional deep learning model (82.11% vs 82.93%, p=1.000). Conclusion In this study, an artificial intelligence diagnostic system based on multi-feature fitting was constructed for the diagnosis of gastric whitish tumorous lesions under white light with good accuracy, which is expected to be used in the future to assist endoscopists clinically to improve the detection rate of gastric whitish tumorous lesions.