Validity, Reliability and Feasibility of an Automated Photographic Measurement/Assessment of Food Intake in the Hospitalized Elderly

Validity, Reliability and Feasibility of an Automated Photographic Measurement/Assessment of Food Intake in the Hospitalized Elderly

Malnutrition affects 50% to 70% of hospitalized elderly people, and is all the more worrying in the elderly because of its clinical impact. A measurement of food consumption is essential to recognize needs, monitor the nutritional status of the elderly in hospital and implement specific therapeutic action such as supplements or an increase in energy-protein to combat malnutrition or the risk of malnourishment. Unfortunately, this measure is rarely done effectively in practice, keeping the patient in nutritional deficit, contributing to a risk of increased morbidity and mortality.

Although weighing food intake is the reference method, it is a routine burden for healthcare teams. To overcome these constraints in hospital environments, intake is estimated by food readings over three consecutive days using a semi-quantitative method. It should be noted that this method remains complex, imprecise and reserved only for the most malnourished patients. In recent years, the development of photographic methods has become an interesting alternative to the measurement by weight. Based on photographs taken before and after the meal in order to deduce what is actually ingested, these methods obtain results comparable to the weighing method, though there is still a number of limitations (need for human intervention, constraint to have standardized menus in weight and lack of nutritional management adapted to patients). To overcome these limitations, an automated photographic method based on modern techniques for automatic processing of 2D and 3D images coupled with techniques derived from artificial intelligence has recently been developed in the investigator's unit, but has not yet been validated.

The originality and innovation of this project lies in the automated analysis of the photos taken and the conversion into percentage of remaining food thanks to the design of algorithms for image preprocessing and neural classification by a 2D and 3D software (patent pending).

Inclusion Criteria:

• patient who has given oral consent to participate
• adult patient
• inpatient geriatric rehabilitation follow-up care (SSRG) and acute geriatric units
• patient eating alone or with help

Exclusion Criteria:

• patient with enteral or parenteral nutrition
• patient not affiliated to a social security scheme
• end-of-life patient or palliative care