<
Enterprise AI Analysis: AI powered dietary proportion assessment for improving accuracy and practicality of the balanced meal plate model
>
Scientific Reports, Article in Press (2025)
AI powered dietary proportion assessment for improving accuracy and practicality of the balanced meal plate model
Authored by Worasit Choochaiwattana, Patinya Jaruariyanon, Assaree Jitpranee, Rawirin Deecharoen, Thanaporn Kaewpradup, Charoonsri Chusak & Sirichai Adisakwattana
Executive Impact: AI-Driven Dietary Health Innovation
This study introduces an AI-based application designed to enhance the accuracy and practicality of the 2:1:1 balanced meal plate model for dietary assessment. It demonstrates superior accuracy compared to human experts (dietetics students and registered dietitians) in estimating food proportions for complex Thai dishes like Hainanese Chicken Rice and Shrimp Paste Fried Rice, achieving significantly lower mean absolute errors (MAE). While performance was comparable for Egg Noodle due to visual complexity, user satisfaction surveys revealed moderate to high approval for the application's speed, ease of use, and aesthetic appeal. Participants recognized its potential for nutrition education, dietary assessment, and clinical application, indicating a strong positive attitude towards its utility and a willingness to recommend it. The tool holds significant promise for promoting healthier eating habits, aiding weight management, and supporting dietary modifications for conditions like diabetes and cardiovascular disease by providing real-time visual feedback and simplifying portion control.
- ✓ AI significantly outperforms human experts in assessing dietary proportions for culturally specific, complex Asian dishes.
- ✓ The application effectively addresses limitations of traditional dietary assessment by providing objective, accurate proportion estimates.
- ✓ High user satisfaction and positive attitudes highlight the tool's practical utility for dietetics students, registered dietitians, and public health education.
- ✓ The AI model enhances adherence to the 2:1:1 balanced meal plate model, fostering healthier eating habits and supporting weight management.
- ✓ This research establishes a foundation for AI-driven nutrition tools that are adaptable to diverse culinary contexts, overcoming previous limitations of existing calorie-focused systems.
0.6%
MAE Reduction for Grains/Starches (AI vs Human, Shrimp Paste Fried Rice)
61%
Users Reported Moderate Overall Satisfaction
91%
Users Satisfied with Data Processing Speed
57%
Found Application Practical for Dietetics
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
AI Outperforms Human Experts
3.2%
Average MAE for Grains/Starches (AI) vs 4.9% (RD) on Hainanese Chicken Rice (p=0.017)
The AI model demonstrated significantly lower mean absolute error in estimating grains/starches for Hainanese Chicken Rice compared to Registered Dietitians, indicating superior accuracy in portion assessment for this complex dish.
Comparative Accuracy: AI vs. Human Estimators
| Category | AI Performance | Human Performance (ND/RD) |
|---|---|---|
| Hainanese Chicken Rice |
|
|
| Shrimp Paste Fried Rice |
|
|
| Egg Noodle |
|
|
The AI model consistently demonstrated superior accuracy for Hainanese Chicken Rice and Shrimp Paste Fried Rice, while showing comparable performance to human experts for Egg Noodle, highlighting challenges with visually complex dishes where ingredient colors overlap.
Addressing Complexity: The Egg Noodle Challenge
Case Study: Challenges with Complex Dishes
The study revealed that for dishes like Egg Noodle, where food components such as bean sprouts, egg noodles, and fish balls share similar colors and textures, the AI's performance was comparable to human estimators, rather than superior. This finding aligns with previous research on AI food recognition tools and highlights the need for further model refinement to better distinguish overlapping ingredients and handle visual complexities in diverse culinary contexts. This informs future development efforts to enhance the AI's ability to accurately assess such dishes.
No Significant Difference
Performance was on par with human experts for Egg Noodle components due to similar ingredient colors and textures, presenting a key area for future AI model enhancements.
AI Model Development & Evaluation Process
Enterprise Process Flow
Image Collection (3 Thai Dishes, 3 Portion Variations)
→
Pixel Labeling (Grains, Proteins, Vegetables)
→
Data Augmentation (Rotation, Brightness, etc.)
→
U-Net/VGG16 Model Training
→
Performance Evaluation (MIoU, MAE)
AI Model Training Dataset
3
Thai Dishes Used for Training
15
Base Images per Dish
225
Total Images per Dish in Final Dataset
68
Dataset Split for Training
The AI model was trained on a carefully curated dataset of popular Thai dishes, using pixel-level annotations and data augmentation to ensure robust learning for accurate dietary proportion assessment.
AI Enhances Dietary Adherence to 2:1:1 Model
61%
Users recognized AI as a valuable resource for nutritional education
Over half of the participants agreed that the AI tool shows potential as a practical resource for nutrition education and dietary assessments, promoting adherence to the 2:1:1 model and facilitating healthy lifestyle changes.
Key User Satisfaction Metrics
91
Satisfied with Speed of Data Processing
61
Overall Moderate Satisfaction with Accuracy
78
Satisfied with Aesthetic Appeal
61
Very Satisfied with Ease of Use
User feedback indicated high satisfaction with the technical aspects and usability of the AI application, alongside a generally positive view of its accuracy for dietary proportion assessment.
Calculate Your Potential AI ROI
Estimate the time savings and financial benefits AI could bring to your organization's dietary assessment and nutritional management processes.
Annual Cost Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A phased approach to integrating AI for enhanced dietary assessment and management within your enterprise, leveraging the insights from this research.
Phase 1: Discovery & Strategy
Conduct a detailed assessment of current dietary assessment workflows, identify key challenges, and define specific objectives for AI integration. Establish success metrics and a clear project scope based on organizational needs.
Phase 2: Data Preparation & Model Customization
Curate and annotate relevant dietary image datasets, potentially expanding on existing models with your organization's specific food items. Customize and train the AI model (e.g., U-Net with VGG16 backbone) to optimize accuracy for your unique culinary context.
Phase 3: Pilot Deployment & User Training
Implement a pilot version of the AI application within a controlled environment. Train dietitians, nutritionists, or relevant staff on its use, gathering initial feedback on usability and practical application.
Phase 4: Iterative Refinement & Full Rollout
Based on pilot feedback, refine the AI model and application features. Address any limitations, particularly for complex or culturally specific dishes. Scale the solution across the organization, ensuring comprehensive integration and ongoing support.
Phase 5: Performance Monitoring & Continuous Improvement
Continuously monitor the AI system's performance, user satisfaction, and impact on dietary outcomes. Implement regular updates and retraining cycles to adapt to evolving dietary patterns and improve long-term accuracy and utility.
Ready to Transform Your Nutritional Intelligence?
Schedule a personalized consultation with our AI specialists to explore how these advanced dietary assessment capabilities can be tailored to your enterprise needs.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat