Enhancing Food Intake Tracking in Long-term Care With Automated Food Imaging and Nutrient Intake Tracking (AFINI-T)...

Area of Science:

• Biomedical Engineering
• Nutritional Science
• Artificial Intelligence

Background:

• Malnutrition affects half of long-term care (LTC) residents, increasing health risks and decreasing quality of life.
• Current methods for tracking food intake in LTC are subjective and inefficient.

Purpose of the Study:

• To present an automated food imaging and nutrient intake tracking technology tailored for LTC environments.
• To address the need for objective, repeatable, and environmentally adaptable malnutrition tracking.

Main Methods:

• A needs assessment was conducted with 21 LTC staff members.
• A deep learning model, including a convolutional autoencoder, was trained for food segmentation and classification using overhead images.
• Food intake was quantified by estimating the volume difference between full and leftover plates.

Main Results:

• The system achieved 88.9% top-1 classification accuracy across 12 meal scenarios.
• Nutrient intake estimation by volume showed strong linear correlation (r²=0.92-0.99) with mass-based estimates.
• The automated method demonstrated good agreement with the gold-standard weighed food method.

Conclusions:

• The developed deep learning system offers a feasible and accurate method for computational nutrient sensing.
• This technology can provide objective tracking of food intake, aiding in malnutrition prevention and management in LTC.
• Positive end-user engagement suggests practical applicability in real-world LTC settings.