AI System for Estimating Leftover Hospital Food
Author Information
Author(s): Mavragani Amaryllis, Sure Tharun Anand Reddy, Kommireddy Shreeven, Aljerf Loai, Tagi Masato, Hamada Yasuhiro, Shan Xiao, Ozaki Kazumi, Kubota Masanori, Amano Sosuke, Sakaue Hiroshi, Suzuki Yoshiko, Konishi Takeshi, Hirose Jun
Primary Institution: Tokushima University Graduate School
Hypothesis
Can an AI model accurately estimate leftover liquid food in clinical environments compared to visual estimation methods?
Conclusion
The AI-based food intake estimation system showed a high correlation with actual values and was more accurate than image visual estimation, but less accurate than direct visual estimation.
Supporting Evidence
- The AI estimation had a mean error of 0.61, indicating a 6.1% error rate.
- The AI estimation showed a smaller RMSE compared to image visual estimation.
- Spearman rank correlation coefficients for AI estimation were high (ρ=0.89-0.97).
- The AI estimation was validated against actual measurements using the weighing method.
Takeaway
Researchers created a smart system that uses AI to guess how much liquid food patients leave behind, and it works better than just looking at the food.
Methodology
The study compared AI estimations of leftover food with visual estimations by dietitians and nurses, using 300 dishes of liquid food.
Potential Biases
Potential bias in visual estimations due to differences in training and experience among staff.
Limitations
The study was conducted at a single facility, and the accuracy may vary with different menus and types of food.
Participant Demographics
Patients in a hospital setting receiving liquid food.
Statistical Information
P-Value
0.82
Statistical Significance
p<0.001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website