An Artificial Intelligence-Based System to Assess Nutrient Intake for Hospitalised Patients

Regular monitoring of nutrient intake in hospitalised patients plays a critical role reducing the risk disease-related malnutrition. Although several methods to estimate have been developed, there is still clear demand for more reliable and fully automated technique, as this could improve data accuracy reduce both burden on participants health costs. In paper, we propose novel system based artificial intelligence (AI) accurately intake, by simply processing RGB Depth (RGB-D) image pairs captured before after meal consumption. The includes multi-task contextual network food segmentation, few-shot learning-based classifier built limited training samples recognition, an algorithm 3D surface construction. This allows sequential estimation consumed volume, permitting automatic each meal. For development evaluation system, dedicated new database containing images recipes 322 meals assembled, coupled annotation using innovative strategies. Experimental results demonstrate that estimated highly correlated (>0.91) ground truth shows very small mean relative errors (<20%), outperforming existing techniques proposed assessment.