A/Professor Nazarbayev University, School of Medicine Astana, Astana, Kazakhstan
Objectives: Monitoring dietary intakes is critical for assessing population diet-related health and nutritional status, providing essential data to inform public health nutrition interventions. A significant challenge in this process is the accurate estimation of food portion sizes, which often leads to gaps in dietary assessments. This study addresses this issue by evaluating the potential of integrating computer vision technology into dietary assessments to enhance accuracy of portion estimates, and ease of use for the general population.
One of the research aims is: How accurately can food volumes be estimated from images by trained humans and machine learning/digital tools? This study focuses on portion size estimation for foods and beverages commonly consumed in Central Asia.
Methods: The experimental design involved 30 commonly consumed food items and 10 beverages from Central Asia. Participants were divided into three groups:
1. Trained individuals using visual references (e.g., portion size photographs or utensil sizes), 2. Untrained individuals without visual aids, and 3. Users of a digital tool
Each participant was randomly assigned digital photographs of food and beverages and asked to estimate the portion sizes of meals. These estimations were recorded and compared with the actual weighed portions. Participants also indicated their confidence levels and note any difficulties encountered.
Results: The Wilcoxon signed-rank tests for small and average portions revealed significant discrepancies between actual and estimated values, with the most significant discrepancies observed for smaller portions, underscoring the need to refine estimation methods, particularly for smaller portions. Correlation coefficients will further assess the accuracy of human raters and digital tool estimations.
Conclusions: This study aims to provide valuable insights into effective dietary assessment tools and methods, contributing to public health nutrition advancements in Central Asia. Incorporating digital tools into national dietary surveys could enhance the accuracy of food consumption data, enabling the development of targeted nutrition interventions tailored to the region.
Funding Sources: Nazarbayev University, under the Faculty Development Competitive Research Grant Program (Grant No. 201223FD2603) and the Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan (Grant No. AP23485288)
