Developing the culturally-valid food image dataset (CV-FID): Protocol and application to hedonic food images

Background Food images are widely used as a virtual substitute for real food in ingestive behavior research. Ensuring scientific rigor requires high-quality, standardized, and culturally valid food image datasets (CV-FID) that account for cross-cultural differences in food recognition and preferences. Objective This study presents a protocol for developing a CV-FID and demonstrates its application in creating the Hedonic Food dataset tailored to the Israeli adult population. Methods The CV-FID standardized protocol follows a five-stage process: (1) an expert focus group to select culturally-relevant food items; (2) comprehensive nutritional analysis; (3) standardized professional food photography; (4) a normative rating survey administered to a representative Israeli adult sample; and (5) integration of food images, nutritional data, and normative ratings into a designated database. This protocol is illustrated through the development of a Hedonic Food Image Dataset (HFID). Results The final HFID included 45 culturally-relevant food items, which were captured with high-resolution images, detailed nutritional information, and comprehensive normative ratings from 157 adults (mean age: 41.9 ± 15.1 years; mean BMI: 26.5 ± 6.0 kg/m 2 ). The survey covered recognition rates, frequency of consumption, perceived healthiness, sensory attributes, and hedonic response, with internal validity confirmed with descriptive trends and correlational analyses. Conclusions The CV-FID protocol provides a structured framework for developing a culturally valid, high-quality food image dataset. By adopting this protocol, researchers can ensure greater consistency and applicability when developing future food image databases across diverse settings. The HFID illustrates the protocol's application and serves as a valuable resource for nutrition-related research.