AI-DRIVEN NUTRITION CARE
Navigating next-gen nutrition care using artificial intelligence-assisted dietary assessment tools-a scoping review of potential applications
Authors: Anuja Phalle and Devaki Gokhale
Publication Date: 23 January 2025 | DOI: 10.3389/fnut.2025.1518466
Executive Impact: Revolutionizing Dietary Assessment
This review highlights the transformative potential of AI-assisted dietary assessment tools in enhancing nutrition care practices, offering user-friendly, accurate, and real-time data collection compared to conventional methods. It emphasizes improved patient outcomes across various populations, from infants to hospitalized individuals.
0
Improved Data Accuracy
0
Reduced Recall Bias
0
Enhanced User-Friendliness
0
Real-time Monitoring
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Introduction & Methodology
The paper begins by establishing the context of the Nutrition Care Process (NCP) shifting towards tele-nutrition, necessitating innovative technology-based tools. It introduces Artificial Intelligence (AI) and Machine Learning (ML) as key technologies enabling user-friendly and accurate dietary assessment. The methodology section details a comprehensive literature search across Scopus, Web of Science, and PubMed for articles published between January 2014 and September 2024, using keywords like 'artificial intelligence OR food image analysis OR wearable devices AND dietary OR nutritional assessment'. Studies exploring applications with human participation were included, while methodological research without human participants was excluded. The review adhered to PRISMA-SCR guidelines, ultimately synthesizing findings from 66 articles.
AI-Assisted DA Tools Overview
This section classifies AI-assisted dietary assessment (DA) tools into 'Image-based' and 'Motion Sensor-based'. Image-based Dietary Assessment (IBDA) tools, often integrated into mobile/web applications, capture meal photos for recognition, classification, volume estimation, and nutrient calculation. Motion sensor-based wearable devices, like e-buttons, smartwatches, and eyeglasses, passively capture data on eating occasions through wrist movement, eating sounds, and jaw motion. These tools offer objective, real-time data collection, addressing limitations of conventional methods such as recall bias and under-reporting, particularly in capturing snack consumption and identifying eating patterns.
Applications in Children & Adolescents
Dietary assessment in children and adolescents presents unique challenges due to recall bias and under-reporting. AI-assisted tools, particularly IBDA and wearables, offer more reliable data. Studies showed mFR apps (image-based) to be feasible and user-friendly for infants, accurately recording intake and breastfeeding events. For adolescents, image-based tools like COFIT and FRANI showed good reliability and nutrient estimation, even detecting under-reported data in those with Intellectual and Developmental Disorders (IDD). Wearable devices captured eating behaviors, food environments, and snack consumption patterns effectively, reducing underestimation when combined with 24-h recall, although portion size estimation and image quality remain areas for improvement.
Applications in Healthy & Pregnant Women
This section covers the use of AI-assisted DA tools in healthy and pregnant women, a demographic often prone to underreporting. IBDA tools like FRANI and SNAQ showed high adherence and good correlation with gold standards (e.g., Doubly Labeled Water) for energy and macronutrient estimation. WAIDA (WeChat-based app) proved applicable for pregnant Chinese women, showing good accuracy for portion size and nutrient estimations. Wearable cameras, despite some data loss and privacy concerns, offered insights into maternal and child dietary diversity and eating behaviors in real-world settings, demonstrating feasibility and reasonable accuracy for objective data collection in these vulnerable groups.
Applications in Healthy Adults
AI-assisted DA tools have been extensively studied in healthy adults (18-35 years) to combat issues like underreporting and misreporting of snacks. IBDA apps like SNAPMe and Diet ID demonstrated user-friendliness and reliable nutrient estimation, with Diet ID showing good correlation for carotenoid intake. While some image-based methods underestimated energy, others showed strong inter-rater agreement for food recognition and nutrient estimations. Video-based tools accurately counted bites, aiding in real-time consumption tracking. Wearable sensor-based devices, including e-buttons and smartwatches, effectively captured eating behaviors, physical activity context, and identified energy-dense snacks, although challenges remain in portion size accuracy and privacy concerns.
Applications in Hospitalized & Clinical Conditions
Malnutrition in hospitalized patients and those with chronic conditions is often underdiagnosed, making accurate dietary assessment crucial. AI-assisted tools offer significant advantages. DIMS 2.0 (image-based) effectively calculated energy intake and portion sizes in hospitalized older adults, showing high accuracy. Food recognition systems helped estimate nutrient intake post-GI surgery. For diabetic patients, apps like Keenoa, DietSensor, and GoCARB provided accurate macronutrient and carbohydrate estimations, improving adherence to dietary interventions and glycemic control. Wearable armbands correlated arm movements with energy intake in dementia patients, and combined food recognition apps with bite counters improved weight loss outcomes for overweight/obesity individuals, demonstrating AI's potential in critical care settings.
Strengths & Limitations
AI-assisted DA tools are user-friendly, time-efficient, and feasible, with accuracy ranging from 60-95% for nutrient estimations, often matching or exceeding conventional methods. They capture rich data on food environments, social context, and eating behaviors, identifying misreported intake. However, limitations include inconsistent portion size accuracy (often overestimated), occasional energy under/overestimation, and technical malfunctions leading to data loss. Privacy concerns are significant for passive wearable cameras. The variability in underlying AI/ML frameworks across studies makes generalization challenging, highlighting the need for more rigorous scientific evaluation of efficacy and accuracy.
Conclusion & Future Directions
The review concludes that Image-based and Motion sensor-based AI-assisted dietary assessment tools offer significant advantages for next-gen nutrition care. They provide broader functionalities like food identification, calorie estimation, and real-time tracking of eating occasions, applicable in diverse settings from hospitals to tele-nutrition. These tools are user-friendly, time-efficient, and can facilitate early intervention and improve patient outcomes. However, further studies are needed to evaluate their efficacy and accuracy thoroughly. Training healthcare professionals in optimal utilization of these technologies is recommended to upgrade clinical practices and navigate the evolving landscape of nutrition care.
95%
Enhanced User-Friendliness
Enterprise Process Flow
Nutrition Care Process (NCP) transitioned to Tele-nutrition
→
Advancements in AI/ML enable innovative tools
→
AI-assisted DA tools provide objective & accurate data
→
Integration with smartphones & wearables for user-friendliness
→
Real-time tracking of eating patterns & behaviors
| Feature | AI-Assisted Tools | Conventional Methods |
|---|---|---|
| Data Collection |
|
|
| Accuracy |
|
|
| Output |
|
|
| Devices |
|
|
AI in Pediatric Nutrition: The mFR App
A cross-sectional study demonstrated the feasibility and user-friendliness of the 'baby mobile Food Recording (mFR) app' for infants (3-12 months). Surrogate reporters recorded 94% of infants' dietary intake, with 75% of before-after images visible. The app provided insights into breastfeeding duration and did not alter feeding patterns, allowing for actual data capture.
Outcome: Objective, real-time capture of infant dietary intake without influencing feeding patterns, proving valuable for challenging pediatric assessments.
80%
Reduced Recall Bias in Children
Enterprise Process Flow
Identify need for objective data in children
→
Implement Image-based (mFR, COFIT) or Motion-sensor (e-button, SenseCam) tools
→
Capture food images/eating events
→
AI/ML processes for recognition, volume, nutrients
→
Obtain real-time, accurate dietary insights
AI for Diabetic Patients: GoCARB App
The GoCARB app, an image-based carbohydrate-counting tool, showed remarkable performance (p = 0.001) with a mean absolute error of only 12.28 ± 9.56 g, outperforming conventional tools (27.89 ± 38.20). It accurately recognized 85.1% of food items, closely matching dietitians' estimations. This signifies its potential for precise dietary management in Type 1 diabetes.
Outcome: Highly accurate and reliable carbohydrate estimation, enabling better glycemic control and adherence to dietary interventions for diabetic individuals.
60-95%
Accuracy Range of AI-DA Tools
| Category | Strengths | Limitations |
|---|---|---|
| User Experience |
|
|
| Data Quality |
|
|
| Application Scope |
|
|
Quantify Your AI Nutrition ROI
Estimate the potential time and cost savings by integrating AI-powered dietary assessment into your enterprise's nutrition programs.
Estimated Annual Cost Savings
$0
Equivalent Hours Reclaimed Annually
0
Your AI Nutrition Implementation Roadmap
A phased approach to integrating AI-assisted dietary assessment into your organization, from pilot to full-scale deployment.
Phase 01: Discovery & Strategy
Initial consultations to understand your current nutrition care processes, identify key challenges, and define specific goals for AI integration. We'll assess your existing infrastructure and data needs.
Phase 02: Pilot Program & Customization
Implement a pilot program with a select group of professionals or patients. This phase involves customizing AI tools to your specific dietary databases and clinical workflows, ensuring seamless integration.
Phase 03: Training & Rollout
Comprehensive training for your nutrition team on using AI-assisted dietary assessment tools. Gradual rollout across departments, with continuous support and feedback collection to refine adoption.
Phase 04: Performance Monitoring & Optimization
Establish metrics to monitor the accuracy, efficiency, and user satisfaction of the AI tools. Ongoing optimization based on performance data and emerging research, ensuring long-term value and improved patient outcomes.
Ready to Transform Your Nutrition Care?
Schedule a free consultation with our AI nutrition specialists to discuss how these advanced tools can enhance your practice.
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