Description

The RICOSALUD1 project aims to research and develop an intelligent system based on 2D/3D computer vision and artificial intelligence for the automatic monitoring of food intake and waste in hospitals. Currently, healthcare professionals do not have objective tools that allow them to accurately determine how much a patient has actually eaten, which makes it difficult to detect risks such as malnutrition or dysphagia early on. This project seeks to address this need by designing an experimental prototype capable of analyzing hospital trays before and after meals.

The proposed solution is based on capturing RGB and depth images (RGB-D technology) to reconstruct the food served and the waste generated after ingestion in three dimensions. Based on these images, advanced artificial intelligence models will be developed that are capable of identifying food, segmenting it correctly even in complex conditions (mixtures, occlusions, lighting variability), and estimating the volume consumed by comparing the initial and final states of the tray. The project also includes the generation of specific datasets, the use of state-of-the-art segmentation models, and innovative techniques such as few-shot learning and synthetic data generation.

With a duration of 12 months and developed in collaboration between the Recoletas Hospital Group and the CARTIF technology centre, the project is limited in this phase to experimental validation in a controlled environment, without clinical integration or use with real patients. The expected impact includes improved nutritional monitoring, optimization of hospital resources, and reduction of food waste, while also aligning with new regulatory requirements on sustainability. RICOSALUD1 thus lays the groundwork for future technological implementation that will enable more accurate, personalized, and efficient nutritional management in the hospital setting.