=Paper=
{{Paper
|id=Vol-3777/paper23
|storemode=property
|title=Development of Intelligent Nutrition Surveillance System with OpenAI API
|pdfUrl=https://ceur-ws.org/Vol-3777/paper23.pdf
|volume=Vol-3777
|authors=Vadym Yevtushenko,Kyrylo Rukkas,Tetyana Chumachenko
|dblpUrl=https://dblp.org/rec/conf/profitai/YevtushenkoRC24
}}
==Development of Intelligent Nutrition Surveillance System with OpenAI API==
https://ceur-ws.org/Vol-3777/paper23.pdf
Development of Intelligent Nutrition Surveillance System
with OpenAI API
Vadym Yevtushenko1, Kyrylo Rukkas2 and Tetyana Chumachenko3
1
National Aerospace University “Kharkiv Aviation Institute”, Vadym Manko str., 17, Kharkiv, 61070, Ukraine
2
V.N. Karazin Kharkiv National University, Freedom sq., 4, Kharkiv, 61001, Ukraine
3
Kharkiv National Medical University, Nauky ave., 4, Kharkiv, 61001, Ukraine
Abstract
Nutrition surveillance is crucial in monitoring public health by identifying dietary deficiencies and tracking
nutritional trends. Traditional systems often struggle with data collection and analysis challenges,
particularly in crises. Artificial intelligence (AI) offers promising solutions for improving the efficiency and
accuracy of such systems. This study aims to develop an intelligent nutrition surveillance system that
leverages the OpenAI API to provide personalized dietary recommendations, improve real-time data
processing, and enhance nutrition management. The system integrates natural language processing (NLP)
and machine learning models for food recognition and nutrient estimation. Data is collected through a user-
friendly interface, and personalized meal plans are generated based on user inputs. The system's
performance was tested using a combination of manual data entry and food image recognition, with user
feedback guiding further refinements. The OpenAI-based system successfully provided real-time,
personalized nutrition plans. The NLP model accurately processed user queries, while the food recognition
model performed well with simple meals and struggled with complex dishes. User satisfaction was generally
high, but some data input guidance and food recognition accuracy improvements were noted. This research
demonstrates the practical application of AI in nutrition surveillance, particularly in resource-constrained
settings. The system's ability to generate personalized recommendations using real-time inputs represents
a significant advancement in public health technology. The study presents a scalable and adaptable
framework for integrating AI into nutrition surveillance, showing strong potential for further application
in individual and population health. Future research should focus on refining image recognition algorithms
and incorporating automated data collection methods to enhance accuracy and applicability.
Keywords
Artificial Intelligence, OpenAI, nutrition surveillance, personal assistance, public health informatics 1
1. Introduction
Nutrition surveillance is critical in monitoring population health, particularly by tracking dietary
intake, identifying nutritional deficiencies, and detecting emerging trends that may indicate broader
public health concerns. Effective nutrition surveillance systems can inform policy decisions, guide
public health interventions, and prevent malnutrition-related illnesses [1]. However, many existing
systems face data collection, processing, and analysis challenges, often relying on self-reported data
or surveys that may lack accuracy and timely insights [2]. These limitations underscore the need for
more intelligent and adaptable approaches to nutrition monitoring, especially in vulnerable
populations.
The problem of nutrition surveillance has become particularly acute during the ongoing full-scale
Russian invasion of Ukraine. War and conflict often disrupt food supply chains, displace populations,
and limit access to essential resources, contributing to food insecurity and poor nutritional outcomes
[3]. In Ukraine, the conflict has severely affected food availability, leading to a rise in malnutrition,
especially in regions with limited humanitarian aid [4]. Existing nutrition surveillance systems,
already strained by the challenges of the COVID-19 pandemic, have been further hindered by
logistical constraints, making it difficult to track and address the nutritional needs of affected
ProfIT AI 2024: 4th International Workshop of IT-professionals on Artificial Intelligence (ProfIT AI 2024), September 25–27,
2024, Cambridge, MA, USA
v.o.yevtushenko@student.khai.edu (V. Yevtushenko); rukkas@karazin.ua (K. Rukkas); tatalchum@gmail.com
(T. Chumachenko)
0009-0007-0733-3034 (V. Yevtushenko); 0000-0002-7614-0793 (K. Rukkas); 0000-0002-4175-2941 (T. Chumachenko)
© 2024 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Wor
Pr
ks
hop
oceedi
ngs
ht
I
tp:
//
ceur
-
SSN1613-
ws
.or
0073
g
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
populations [5]. These challenges necessitate more advanced systems that can operate effectively in
crisis conditions.
The impact of nutrition on public health is well-documented, with poor nutrition being a leading
cause of morbidity and mortality worldwide [6]. In the context of the war in Ukraine, malnutrition
exacerbates pre-existing health conditions and increases vulnerability to disease outbreaks, such as
influenza and other communicable diseases [7]. Adequate nutrition is essential for maintaining
immune function and overall well-being, particularly in war-affected regions where healthcare
systems are overwhelmed and limited access to food [8]. Addressing nutritional needs is crucial for
short-term survival and long-term recovery, highlighting the importance of robust nutrition
surveillance systems in managing public health during conflict.
Artificial intelligence (AI) has emerged as a valuable tool for public health, offering solutions to
enhance data collection, analysis, and decision-making processes. AI can help bridge the gaps in
traditional nutrition surveillance by automating data collection from diverse sources, including social
media, satellite imagery, and mobile health platforms [9]. Machine learning models can process large
datasets to detect patterns in nutritional intake and predict potential public health crises before they
manifest [10]. In conflict zones, AI can support rapid food security and malnutrition assessments by
integrating real-time data, providing governments and humanitarian organizations with actionable
insights to guide interventions [11].
Large language models (LLMs), such as OpenAI’s GPT, have demonstrated significant potential
as personal assistants in public health. These models can assist individuals in making informed
nutritional decisions by offering personalized dietary recommendations based on available data,
preferences, and constraints [12]. In war-torn regions, where access to healthcare professionals may
be limited, LLMs can be an accessible resource for individuals and healthcare workers. Their ability
to analyze vast amounts of data, respond to queries in natural language, and provide tailored
guidance makes them a valuable asset in supporting nutrition surveillance efforts, particularly in
resource-constrained settings.
The study aims to develop an intelligent nutrition surveillance system using OpenAI API.
2. Current research analysis
Data-driven nutrition surveillance systems have gained prominence in recent years as technological
advancements enable more efficient and accurate collection and analysis of nutritional data. These
systems leverage large-scale datasets, including electronic health records, food consumption surveys,
and remote sensing data, to monitor real-time population nutrition trends. Using machine learning
algorithms and data analytics tools, these systems can identify emerging patterns, predict future
nutritional risks, and support targeted public health interventions. Integrating data from diverse
sources such as mobile health applications, social media platforms, and geographic information
systems (GIS) allows for more comprehensive assessments of dietary behaviours and food
accessibility, providing valuable insights into nutritional challenges at the community and individual
levels. However, despite these advancements, current data-driven nutrition surveillance systems still
face challenges related to data quality, accessibility, and interoperability, particularly in low-resource
or conflict-affected settings.
The paper [13] discusses the implementation and impact of a mobile-based nutrition surveillance
system to improve accountability and real-time monitoring of nutritional outcomes in India. The
Poshan Tracker system, part of the country’s Integrated Child Development Services Scheme,
provides transparent data on anthropometric outcomes, food distribution, and the functioning of
Anganwadi Centers (AWCs). Over the study period, there were significant improvements in the
operational consistency of AWCs and the distribution of supplementary food to vulnerable groups.
However, the paper highlights limitations in data accuracy, with discrepancies between Poshan
Tracker data and the National Family Health Survey, potentially caused by differences in
measurement techniques and reporting biases. This indicates the need for further system refinement
to ensure more reliable and consistent data for policy-making and program adjustments.
The paper [14] explores the dietary patterns (DPs) that help protect against hypertension (HTN)
in a nationwide study of Chinese adults. The researchers used reduced rank and partial least square
regression to identify key dietary components of HTN protection. Their findings suggest that diets
high in fresh vegetables, fruits, mushrooms, dairy products, and legumes are associated with lower
odds of HTN. The study highlights the relevance of adapting dietary patterns to local habits, as
traditional Western diets like the DASH diet may not be suitable for Chinese populations. However,
a limitation is the potential lack of generalizability to other ethnic groups due to the focus on Chinese
adults and the specific dietary patterns derived from their habits.
The chapter [15] explores integrating AI technologies into nutrition and fitness to improve health
outcomes. It emphasizes how AI can assist in personalized nutrition and fitness strategies by
analyzing large datasets and adapting recommendations based on individual factors such as genetics,
lifestyle, and environment. AI applications, such as machine learning algorithms and wearable
technologies, are discussed in the context of improving diet, monitoring physical activity, and
supporting rehabilitation. However, the paper highlights limitations in the generalizability of AI
models due to the complexity and variability of human health data, as well as challenges related to
data integration and privacy concerns, which must be addressed for broader implementation.
The paper [16] comprehensively reviews AI nutritionists’ current advancements and applications,
focusing on software designed for dietary monitoring, food recognition, and nutrient
recommendations. The study systematically evaluates 177 AI nutritionists, highlighting their
growing importance in personalized nutrition and health monitoring. While AI nutritionists have
significantly improved dietary tracking and personalized dietary advice, the paper identifies
limitations in the current level of intelligence of these tools, noting that most systems rely on basic
algorithms and have yet to fully exploit advanced AI capabilities, such as deep learning or molecular-
level food behaviour prediction.
The paper [17] presents a novel system designed to estimate the nutritional content of meals
served in compartment trays using image-based technology. The platform employs a depth camera
and Raspberry Pi to capture images and depth data, which are then processed by an algorithm that
integrates dish recognition, portion-size estimation, and nutrition analysis. Using instance
segmentation models like CenterMask with VoVNetV2-99, the system accurately identifies food
items and calculates their nutritional value. The study demonstrates that this AI-driven method
significantly outperforms dietitians in speed and achieves comparable accuracy in calorie estimation.
However, a limitation of the platform is that its performance heavily depends on the quality of the
depth information, and further refinement is needed to handle more complex and varied meal types.
The paper [18] explores the development of an AI-powered mobile application designed to
provide personalized dietary recommendations based on individual user data, such as body mass
index (BMI), age, gender, and food preferences. By utilizing machine learning algorithms, the app
generates tailored diet plans that mimic the role of a nutritionist, helping users maintain a healthy
diet without requiring professional consultation. The system processes user input to recommend
suitable meal plans, saving time and making dietary guidance more accessible. However, the paper
notes limitations in the model’s ability to account for complex medical conditions and dietary
restrictions, which may affect the accuracy of its recommendations. The system also requires further
validation in real-world settings to improve its practical applicability and adaptability to diverse user
needs.
The paper [19] presents a system that leverages artificial intelligence to provide tailored dietary
recommendations and optimize meal planning, particularly in medical settings. The proposed system
uses food images taken before and after consumption to estimate nutrient intake through image
analysis and segmentation techniques to accurately identify food portions. This approach aims to
streamline nutritional assessments, replacing traditional, time-consuming methods often relying on
memory and interviews. However, the system faces limitations, such as variability in food
composition databases, which can lead to inconsistencies in nutrient estimates and challenges in
handling various food types with differing nutritional profiles. Further validation and refinements
are needed to improve the system’s accuracy and reliability across diverse populations and dietary
needs.
In summary, current research in nutrition surveillance has made significant strides in integrating
artificial intelligence, machine learning, and image recognition technologies to improve dietary
assessments and personalized nutrition. Studies have demonstrated the potential of AI-based systems
to provide more accurate and timely nutritional evaluations compared to traditional methods.
However, despite advancements in segmentation techniques, nutrient estimation, and menu
planning, limitations persist in the variability of food composition databases, the need for extensive
manual data input, and the handling of complex medical conditions or diverse populations. These
challenges highlight the importance of developing more robust and adaptable systems. The proposed
study aims to address these limitations by leveraging the capabilities of the OpenAI API to create a
more intelligent and user-friendly nutrition surveillance system. By integrating natural language
processing and advanced machine learning algorithms, this system seeks to provide accurate, real-
time dietary guidance, streamline data collection, and enhance the overall functionality of existing
nutrition surveillance tools.
3. Materials and methods
The intelligent nutrition surveillance system developed in this research integrates various advanced
technologies, with the OpenAI API serving as the core component. The system architecture is built
around a combination of artificial intelligence techniques, including natural language processing
(NLP), machine learning, and image recognition. These technologies provide personalized dietary
recommendations based on user data, ensuring the system can efficiently process, analyze, and
present nutritional information in real time. Figure 1 presents the system’s architecture.
Figure 1: System’s architecture
The system relies heavily on the OpenAI API to generate tailored dietary advice. User inputs,
such as age, gender, weight, height, and dietary preferences, are collected through a user-friendly
interface developed using C# and the Windows Presentation Foundation platform. These inputs are
processed by the OpenAI model, which generates personalized meal plans and nutritional
recommendations. MongoDB Atlas, a cloud-based NoSQL database, securely stores and manages
user data, allowing for scalability and high system availability. The database stores demographic
information, user dietary preferences, and meal history, ensuring personalized recommendations are
based on accurate and comprehensive data.
The system also incorporates a food image analysis component, allowing users to upload photos
of their meals. These images are analyzed using convolutional neural networks (CNNs) for food
identification and segmentation. The CNN model segments the food items from the image, identifies
the types of food, and estimates portion sizes. These segmented food items are then cross-referenced
with a food composition database to calculate the nutritional content, including calories and
macronutrient breakdowns. This data is stored in the user’s profile for future reference and further
refinement of recommendations.
OpenAI API integration allows the system to provide real-time responses to user queries, offering
personalized dietary suggestions. When a user submits a query, the system formulates a request in
JSON format, which includes relevant user data and dietary goals. This request is sent to the OpenAI
API, which processes the input and generates a text-based recommendation or meal plan. The system
then formats this response and presents it to the user in an intuitive and accessible manner through
the interface. The model’s ability to understand and process natural language enables it to provide
contextually relevant advice, even for complex dietary questions.
Image recognition technology is employed for nutrient estimation to enhance the system’s utility.
The system can provide an accurate breakdown of the user’s nutrient intake by analyzing the volume
and type of food in a meal. The food recognition model is trained on a large dataset of labelled food
images to ensure high accuracy. It uses image segmentation techniques to divide the meal into
individual food items, which are then analyzed for nutritional content. This process allows users to
receive real-time feedback on their nutrient intake, making the system a valuable tool for continuous
dietary monitoring.
Figure 2 presents the classes diagram.
Figure 2: The class diagram
The system’s AI models, particularly the GPT model from OpenAI and the CNN for image
analysis, are fine-tuned to improve accuracy and performance in the context of nutrition. The GPT
model is adapted to respond with accurate dietary advice by training it on a diverse range of
nutritional literature and dietary guidelines. Similarly, the food image recognition model undergoes
optimization to handle diverse food types and complex meals, improving the precision of portion
size and nutrient estimation.
Extensive testing and validation were conducted to ensure the system’s reliability and accuracy.
The system’s performance was tested in various conditions, including high user demand, to ensure
it could provide quick and accurate responses. The accuracy of the nutritional recommendations was
evaluated by comparing them with established nutritional standards and expert dietary advice.
Additionally, dietitians and healthcare professionals provided feedback on the system’s interface and
functionality, which was used to refine the user experience further and improve the accuracy of the
generated recommendations.
This study’s methodology establishes a robust framework for an intelligent, AI-powered nutrition
surveillance system that can provide personalized dietary guidance and continuous monitoring,
leveraging the capabilities of OpenAI’s advanced models.
4. Use cases
Tables 1-5 presents the use cases of the system.
Table 1
Use case “Authorization”
Parameter Value
Actors Guest
Goal The application needs to be opened
Main flow 1. The user reaches the authentication page.
2. The user enters a username and password.
3. The system checks the username and password. If the username or
password is incorrect, the alternative flow A1 is executed. If the user enters
correct data, the alternative flow A2 is executed. The use case is then
completed.
Result The guest logs into the system as an authenticated user.
Alternative flow 1 "Username or password is incorrect"
1. The system notifies the user about incorrect data.
2. The system prompts the user to re-enter the data.
3. The use case is completed.
Alternative flow 2 "Correct data"
1. The system notifies the user of successful authentication.
2. The system redirects the user to the main page.
3. The use case is completed.
Postcondition The user is redirected to the main page of the application.
Table 2
Use case “Registration”
Parameter Value
Actors Guest
Goal Register
Main flow 1. The user reaches the authentication page.
2. The user clicks on the "Sign up" button.
3. The user enters a username, a new email, and a password and confirms the
password.
4. The system checks whether the entered data is correct. If it is, alternative
flow A1 is executed; if it is incorrect, flow A2 is executed.
5. After completing alternative flow A1, the system checks if the entered email
is unique. If it is, alternative flow A3 is executed; if a user with the same email
or username already exists, alternative flow A2 is executed.
Result The guest is registered in the system as a new user.
Alternative flow 1 "The entered data is correct."
1. The system begins checking if the user already exists.
2. The use case is completed.
Alternative flow 2 "The entered data is incorrect."
1. The system notifies the user about the incorrect data.
2. The system prompts the user to re-enter the data.
3. The use case is completed.
Alternative flow 3 "No users with the same email or username found in the system."
1. The system notifies the user of successful registration.
2. The system redirects the user to the authentication page.
3. The use case is completed.
Postcondition The user is redirected to the authentication page.
Table 3
Use case “Adding and updating the information of user”
Parameter Value
Actors User
Goal Upload file
Precondition The user must be registered in the web application.
Main flow 1. Click on the "Your details" button.
2. Fill in the data.
3. After completing the data entry, click the "Save" button.
4. The use case is completed.
Result The user's data is uploaded to the database. The system notifies the user of
the successful data saving.
Alternative flow 1 "The user did not select one or more required fields"
1. The user receives an error message.
2. The unfilled fields are highlighted in red.
3. The system prompts the user to re-enter the data.
4. The use case is completed.
Postcondition The user's data is updated on the screen.
Table 4
Use case “Menu generation for healthy diet for the week”
Parameter Value
Actors User
Goal Menu generation
Precondition The user must be registered.
Main flow 1. The user navigates to the generation page by clicking the appropriate
button.
2. The user selects their details if they have not already entered them on the
"Your details" page.
3. The user clicks the "Generate" button.
4. If any data is missing, alternative flow A1 is executed. If all data is
complete, the generation process begins.
5. Once the generation is complete, the user is presented with the generated
healthy diet menu, and an image corresponding to the menu is generated
and set as the background.
6. At the end, the user has the option to regenerate the menu (return to step
5), change the data (return to step 2), or save the menu to favorites
(alternative flow A2 is executed).
7. The use case is completed.
Result The menu is added to the user's favourites.
Alternative flow 1 "Some data is missing"
1. The user receives an error message.
2. The system prompts the user to re-enter the missing data.
3. The use case is completed.
Alternative flow 2 "Saving the menu to favorites"
1. The system notifies the user that the data was successfully saved.
2. The use case is completed.
Postcondition The generation page reopens.
Table 5
Use case “Viiew favorite menus”
Parameter Value
Actors User
Goal View favorite menu
Precondition The user must be registered.
Main flow 1. The user navigates to the favorites page by clicking the appropriate
button.
2. A grid of generated images corresponding to the saved menus is displayed
on the page.
3. By clicking on an image, the text of the menu is opened.
4. The use case is completed.
5. Results
The intelligent nutrition surveillance system successfully provided personalized dietary
recommendations and effectively managed user data. The system’s performance was assessed based
on the quality of personalized recommendations, the accuracy of food recognition, and user feedback.
The OpenAI API demonstrated robust performance in generating meal plans tailored to individual
user profiles. The generated recommendations adhered closely to standard dietary guidelines and
were well-suited to user-specific inputs such as age, weight, and health goals. The personalization
component was highly effective, offering relevant suggestions for users aiming for weight loss,
muscle gain, or balanced nutrition. Validation against established nutritional practices showed that
the system provided nutritionally sound meal plans, confirming the effectiveness of AI in
personalized dietary planning. Figure 3 presents the system’s generation page.
In the food image recognition component, CNNs identified simple food items and estimated their
portion sizes well. The model achieved high accuracy when dealing with individual food items,
making it reliable for common meals. However, it faced challenges when presented with mixed or
complex dishes, where segmentation became less reliable. This limitation reduced the precision of
nutrient estimation for meals containing multiple ingredients. The overall accuracy of the image
recognition was satisfactory, but the need for improvement in handling complex meals was evident.
Figure 4 presents the generated menu.
Figure 3: System’s generation page
Figure 4: Generated menu
User feedback was critical in evaluating the system’s interface and usability. Users appreciated
the straightforward design and found it easy to navigate, particularly the process of inputting
personal information and generating meal plans. While most users were satisfied with the interface,
some suggested adding more detailed options for dietary preferences, such as specific nutrient goals
or food restrictions, to further improve customization. Nutrition experts involved in the testing also
confirmed that the system’s recommendations aligned with accepted dietary practices.
The system’s data management, facilitated by MongoDB Atlas, was efficient and scalable. It
handled large volumes of user data without performance issues, and retrieving stored information
for generating meal plans was seamless. Security protocols ensured data protection, and no data
breaches or security incidents were observed during testing. The system performed well under
different user loads, maintaining stable operation despite increasing data demands. Figure 5 presents
the saved menu.
Figure 5: Saved menu
The system successfully fulfilled its primary objectives. While the image recognition component
requires further refinement for complex meals, the personalized nutrition recommendations and data
management aspects performed effectively. Overall, the system shows strong potential for broader
application in personalized nutrition monitoring.
6. Discussion
The development of the intelligent nutrition surveillance system using OpenAI API marks a
significant contribution to the evolving field of personalized healthcare, specifically in nutrition
monitoring and intervention. This system integrates state-of-the-art artificial intelligence
technologies, offering real-time, data-driven dietary recommendations tailored to individual user
profiles. The ability to automate the generation of meal plans based on user-specific inputs such as
health goals, dietary preferences, and demographic data sets this system apart from traditional
methods, which rely on static guidelines or require professional input. This innovation not only
enhances accessibility to personalized nutrition but also opens new possibilities for improving public
health outcomes, particularly in under-resourced or crisis-affected regions.
The practical novelty of this system lies in its combination of several cutting-edge technologies.
By leveraging OpenAI’s NLP capabilities, the system can engage users conversationally, making
complex nutritional advice easy to understand and apply. This feature is especially important for
individuals who may not have access to dietitians or nutritionists, providing them with a reliable
and interactive source of dietary guidance. The system’s user-friendly interface enables seamless
interaction, allowing users to quickly input their data and receive tailored meal plans. Moreover,
integrating machine learning algorithms for food recognition offers another layer of personalization
by analyzing meal images to estimate nutritional content, creating a comprehensive nutrition
surveillance tool.
However, the system’s limitations must also be acknowledged. The accuracy of the food
recognition component, which relies on CNNs, is limited when confronted with complex or mixed
dishes. In such cases, the model struggles to segment food items accurately, leading to errors in
portion size estimation and nutrient calculation. This presents a challenge, particularly for users with
diverse diets or those consuming culturally specific meals that are harder to categorize. Addressing
this limitation would require improving the model’s ability to handle more intricate food patterns
and increasing the training dataset’s diversity to enhance its applicability across various populations.
Another limitation is the reliance on user-provided data. The effectiveness of personalized dietary
recommendations is contingent upon the accuracy and completeness of the data users input. Errors
such as incorrect weight, height, or dietary goals can skew the system’s outputs, leading to
suboptimal or misleading nutritional advice. Although the system prompts users to correct missing
or inaccurate data, human error remains an inherent challenge, particularly in a self-guided platform.
In the future, integrating more automated data collection methods, such as wearable health devices
that track physical activity and biometrics, could mitigate this issue by reducing the dependence on
manual inputs.
Despite these challenges, the system presents clear opportunities for enhancing individual health
outcomes and broader public health monitoring. One of its most promising applications is in conflict-
affected or resource-poor regions with limited access to healthcare professionals. By automating the
provision of dietary advice, this system can help address malnutrition and related health issues,
offering a scalable solution to global nutrition challenges. The flexibility of the OpenAI API allows
the system to be continuously updated with the latest nutritional research and guidelines, ensuring
that recommendations remain relevant and scientifically grounded.
In addition to public health applications, the system’s design opens doors for personal health
management, where individuals can use the platform to maintain dietary goals, manage chronic
conditions such as diabetes or hypertension, and improve overall well-being. As precision nutrition
grows, systems like this could play a central role in personalizing healthcare, allowing individuals to
adjust their diets based on genetic, lifestyle, and environmental factors.
Looking forward, there are several areas where the system could be enhanced to maximize its
impact. Expanding the database of food images to include a wider variety of dishes from different
cultures would improve the accuracy of the food recognition model, making the system more
adaptable to diverse populations. Additionally, incorporating more advanced AI models, such as deep
learning techniques for food behaviour prediction, could refine the precision of nutrient estimation
and dietary recommendations. Another potential development could involve integrating AI-driven
feedback loops where the system learns from users’ dietary habits over time, enabling even more
personalized and adaptive recommendations.
The intelligent nutrition surveillance system demonstrates the powerful potential of AI to
revolutionize personalized nutrition and public health. While there are limitations related to food
recognition and reliance on user-provided data, these challenges do not diminish the system’s
broader value. The practical novelty of combining AI-driven meal generation, image-based
nutritional analysis, and conversational interfaces allows for significant advancements in both
individual and population health management. With further refinement and expansion, this system
could become an indispensable tool in promoting healthier lifestyles and addressing malnutrition on
a global scale.
7. Conclusions
This study contributes significantly to nutrition surveillance by demonstrating the potential of
integrating artificial intelligence into personalized dietary management, particularly through the
OpenAI API. The system offers a novel approach by generating real-time, tailored nutritional
recommendations based on user data and dietary goals. Its ability to combine natural language
processing for user interaction with machine learning algorithms for food recognition and nutrient
estimation enhances both the accuracy and accessibility of dietary guidance. This approach addresses
the limitations of traditional methods, which often rely on static guidelines and manual input. It
represents a step forward in improving the precision and responsiveness of nutrition monitoring
systems.
One of this study’s key contributions is its practical application in resource-constrained or crisis-
affected settings. By automating dietary recommendations, the system can provide essential support
where access to professional healthcare services is limited. Moreover, the use of AI to analyze user
data and predict nutritional needs opens new avenues for scaling personalized healthcare solutions
globally, benefiting both individual users and public health initiatives.
However, the system also highlights several areas for future research. Enhancing the accuracy of
the food recognition component, particularly for complex or mixed dishes, remains an important
task. Future research should explore the integration of more sophisticated image recognition
algorithms and a broader dataset of food images to improve the system’s applicability across diverse
populations. Additionally, reducing reliance on manual data input by incorporating automated
health data collection, such as from wearable devices, could improve the system’s accuracy and ease
of use.
Looking ahead, expanding the system’s capabilities to include adaptive learning based on user
behaviour and feedback could further personalize dietary recommendations. The system could
evolve into an even more powerful tool for long-term health management by continuously refining
its suggestions based on real-time user data and health outcomes. Further investigation into the
ethical implications of AI in nutrition, particularly regarding data privacy and consent, will also be
essential as such systems become more widely adopted.
This study offers a foundational framework for using AI in personalized nutrition and sets the
stage for future advancements in the field. By addressing current limitations and exploring new
directions in AI-driven nutrition surveillance, future research can build on these findings to create
more effective, accurate, and accessible tools for personal and public health.
Acknowledgements
The study was funded by the Ministry of Health of Ukraine in the framework of the research project
0123U100184 on the topic “Analysis of the impact of war and its consequences on the epidemic
process of widespread infections on the basis of information technologies”.
References
[1] S. Downer, S. A. Berkowitz, T. S. Harlan, D. L. Olstad, and D. Mozaffarian, “Food is medicine:
actions to integrate food and nutrition into healthcare,” The BMJ, vol. 369, p. m2482, Jun. 2020,
doi: 10.1136/bmj.m2482.
[2] J. de Toro-Martín, B. Arsenault, J.-P. Després, and M.-C. Vohl, “Precision Nutrition: A Review
of Personalized Nutritional Approaches for the Prevention and Management of Metabolic
Syndrome,” Nutrients, vol. 9, no. 8, p. 913, Aug. 2017, doi: 10.3390/nu9080913.
[3] D. Chumachenko and T. Chumachenko, “Ukraine war: The humanitarian crisis in
Kharkiv,” BMJ, p. o796, Mar. 2022, doi: 10.1136/bmj.o796.
[4] W. L. Filho et al., “How the War in Ukraine Affects Food Security,” Foods, vol. 12, no. 21, pp.
3996–3996, Oct. 2023, doi: 10.3390/foods12213996.
[5] N. Yazbeck, R. Mansour, H. Salame, N. B. Chahine, and M. Hoteit, “The Ukraine–Russia War Is
Deepening Food Insecurity, Unhealthy Dietary Patterns and the Lack of Dietary Diversity in
Lebanon: Prevalence, Correlates and Findings from a National Cross-Sectional Study,” Nutrients,
vol. 14, no. 17, p. 3504, Aug. 2022, doi: 10.3390/nu14173504.
[6] D. J. Taberna, S. Navas-Carretero, and J. A. Martinez, “Current nutritional status assessment
tools for metabolic care and clinical nutrition,” Current Opinion in Clinical Nutrition and
Metabolic Care, vol. 22, no. 5, pp. 323–328, Sep. 2019, doi: 10.1097/MCO.0000000000000581.
[7] J. M. Quinn V et al., “COVID-19 at War: The Joint Forces Operation in Ukraine,” Disaster
Medicine and Public Health Preparedness, vol. 16, no. 5, pp. 1–8, Mar. 2021, doi:
10.1017/dmp.2021.88.
[8] G. B. Gonzales et al., “Nutrition remains a top priority in Ukraine despite the ongoing war: the
fight against trans fats,” BMJ Global Health, vol. 9, no. 9, p. e016301, Sep. 2024, doi:
10.1136/bmjgh-2024-016301.
[9] D. Chumachenko, M. Butkevych, D. Lode, M. Frohme, K. J. G. Schmailzl, and A. Nechyporenko,
“Machine Learning Methods in Predicting Patients with Suspected Myocardial Infarction Based
on Short-Time HRV Data,” Sensors, vol. 22, no. 18, p. 7033, Sep. 2022, doi: 10.3390/s22187033.
[10] A. Mohammadi, et al., “Comparative study of linear regression and SIR models of COVID-19
propagation in Ukraine before vaccination,” Radioelectronic and Computer Systems, vol. 3, no. 3,
pp. 5–18, Oct. 2021, doi: 10.32620/reks.2021.3.01.
[11] A. Ahmad et al., “AI can empower agriculture for global food security: challenges and prospects
in developing nations,” Frontiers in artificial intelligence, vol. 7, p. 1328530, Apr. 2024, doi:
10.3389/frai.2024.1328530.
[12] D. Chumachenko, O. Sokolov, and S. Yakovlev, “Fuzzy recurrent mappings in multiagent
simulation of population dynamics systems,” International Journal of Computing, vol. 19, no. 2,
pp. 290–297, Jun. 2020, doi: 10.47839/ijc.19.2.1773.
[13] L. M. Jaacks, A. Awasthi, and A. Kalra, “India’s Poshan Tracker: data-driven tool for maternal
and child nutrition,” The Lancet Regional Health - Southeast Asia, vol. 25, pp. 100381–100381,
Mar. 2024, doi: 10.1016/j.lansea.2024.100381.
[14] Y. Yang et al., “Comparison of data-driven identified hypertension-protective dietary patterns
among Chinese adults: based on a nationwide study,” European Journal of Nutrition, vol. 62, no.
7, pp. 2805–2825, Jun. 2023, doi: 10.1007/s00394-023-03195-9.
[15] M. H. B. de Moraes Lopes, D. D. Ferreira, A. C. B. H. Ferreira, G. R. da Silva, A. S. Caetano, and
V. N. Braz, “Use of artificial intelligence in precision nutrition and fitness,” Artificial Intelligence
in Precision Health, pp. 465–496, 2020, doi: 10.1016/b978-0-12-817133-2.00020-3.
[16] Y. Liang et al., “AI nutritionist: Intelligent software as the next generation pioneer of precision
nutrition,” Computers in Biology and Medicine, vol. 178, p. 108711, Aug. 2024, doi:
10.1016/j.compbiomed.2024.108711.
[17] S.-C. Huang, W.-C. Chiang, Y.-T. Yang, and J.-S. Wang, “An Image-based AI Nutrition Analysis
Platform for Food in Compartment Trays,” Proceedings - 2023 14th IIAI International Congress on
Advanced Applied Informatics, pp. 373–374, Jul. 2023, doi: 10.1109/iiai-aai59060.2023.00079.
[18] A. D. Murumkar, A. Singh, B. R. Chachar, P. D. Bagade, and G. Zaware, “Artificial Intelligence
(AI) based Nutrition Advisorusing an App,” International Conference on Sustainable Computing
and Smart Systems, ICSCSS 2023 - Proceedings, pp. 586–590, Jun. 2023, doi:
10.1109/icscss57650.2023.10169703.
[19] K. Azzimani, H. Bihri, A. Dahmi, S. Azzouzi, and M. E. H. Charaf, “An AI Based Approach for
Personalized Nutrition and Food Menu Planning,” 2022 IEEE 3rd International Conference on
Electronics, Control, Optimization and Computer Science, ICECOCS 2022, pp. 1–5, Dec. 2022, doi:
10.1109/ICECOCS55148.2022.9983099.