SAP Health and Policy
SAP Health and Policy
Review Article 2025 Vol. 4 shp2025236

AI for autonomous health care on diabetes diagnostics

Hari Kiran Vege1 , Sri Kamal Yandamuri2 , Jetti Vennela2 , Sai Venkat2
1Additional Dean Academics, Koneru Lakshmaiah Educational Foundation. Vijayawada, India.
2Computer Science and Engineering Koneru Lakshmaiah Educational Foundation. Vijayawada, India.

https://doi.org/10.56294/shp2025236

PDF XML

Abstract

The project aims to improve diabetes prediction using Artificial Intelligence and Machine Learning (AIML) technologies. Diabetes is a chronic disease that needs to be detected early and monitored regularly. Conventional diagnostic methods are based on clinical evaluation and laboratory tests, which are time-consuming and expensive. The system uses cloud computing and machine learning algorithms to create a scalable and effective diabetes prediction model. With patient health data like glucose levels, BMI, age, and insulin levels, the system implements machine learning techniques like Logistic Regression, Random Forest, and Neural Networks to estimate the probability of diabetes. Integration with the cloud provides real-time analytics, data security, and easy access to healthcare professionals.

Keywords

Diabetes, Artificial Intelligence, Machine Learning, Prediction, Cloud Computing

INTRODUCTION

Diabetes mellitus is a chronic illness found in millions globally, which gives rise to such serious complications like cardiovascular disease, renal failure, and neuropathy. Increasing incidents of diabetes exert a high pressure on the health system, for which it becomes imperative to get an early diagnosis and intervention for reducing its long-term effects. But the common diagnostic practices with clinical tests and blood examinations usually involve visiting a healthcare facility for several times and are hence not very efficient or accessible.(1,2)

Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing the field of medicine by enabling computerized, data-based solutions for disease diagnosis and management. AI and ML technologies can evaluate huge amounts of patient data to detect patterns and relationships that may go unnoticed by conventional means. Using AI-based models, diabetes risk assessment in its early stages can be greatly enhanced, resulting in early intervention and improved patient outcomes.(3,4,5)

Cloud computing is key to enhancing the scalability, efficiency, and accessibility of AI-driven diabetes prediction systems. Through integration with cloud, medical facilities can store and analyse large volumes of data without incurring the cost of high infrastructure. Cloud platforms such as AWS SageMaker and Google Cloud AI facilitate effortless deployment of machine learning models with real-time analytics and remote access for healthcare experts.(6,7,8)

The fusion of AI and cloud computing for diabetes prediction guarantees greater precision, better monitoring of patients, and lower costs in healthcare. Such systems permit real-time updating according to novel information, and over time, improve the accuracy of predictive models.(9,10,11) They also make telemedicine solutions possible, supporting remote diagnosis and patient management in underserved areas.(12,13)

This research focuses on creating a cloud-based AIML system that effectively predicts diabetes risk through machine learning algorithms.(14,15) The system uses Logistic Regression, Random Forest, and Neural Networks to analyse patient health records and provide precise risk ratings. Through real-time analytics utilizing cloud services, this system targets enhanced diabetes prediction accuracy while ensuring accessibility and scalability in contemporary healthcare environments.(16,17)

Diabetes mellitus is a long-term condition that affects millions of people across the world and results in grave consequences like cardiovascular disease, renal failure, and neuropathy.(18) While diabetes has been expanding its wings, health systems across the world are encountering mounting pressure to control its long-term effects well. Detection of diabetes at an early stage and timely intervention have emerged as crucial factors in arresting its consequences and improving the patient outcomes.(19,20) Indeed, conventional diagnostic approaches, which are so dependent on medical testing and sequential blood work, have the potential to commit patients to visiting the care center on several occasions. Not only does this lead to inefficiency, but it serves to restrict patient access, most notably for those patients residing on the periphery in distant or medically underserved regions.(21,22)

In the recent past, Artificial Intelligence (AI) and Machine Learning (ML) have begun to revolutionize the practice of medicine, with new and promising fields of disease diagnosis and treatment.(23,24) Through enabling computerized, data-driven solutions, AI and ML technologies can analyze enormous amounts of patient data to discern patterns and connections that are not readily apparent using conventional approaches.(25,26) The use of AI models for the prediction of diabetes risk improves the detection of the disease at an early stage and enables early intervention and potentially the prevention of the severity of long-term complications. By using advanced algorithms, AI systems can deliver quicker, more accurate diagnostic data, enhancing the effectiveness of healthcare delivery and patient care.(27,28)

New advances in Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing the field of medical diagnostics by offering efficient, data-driven solutions. AI and ML technologies can sift through large and complex datasets of patient data to determine patterns, correlations, and risk factors that might not be easily identified through conventional means.(29) Through leveraging these abilities, AI-powered systems can significantly enhance the early detection of diabetes, enabling healthcare professionals to institute timely interventions and preventive measures that can significantly improve patient quality of life.(30)

Application of ML and AI to predict diabetes has the potential to revolutionize the healthcare sector by rendering patient treatment more proactive, efficient, and personalized.(31) As opposed to care initiated after symptoms, AI-enabled systems provide early risk-based interventions, which will reduce the number and severity of diabetes complications immensely. Moreover, application of AI to automate work of diagnosis has the potential to release the burden of healthcare professionals to focus more time on patient treatment and high-level.(32,33)

This research seeks to develop an AI and ML-based diabetes risk prediction system with high accuracy. By applying multiple machine learning techniques to analyze structured patient health records, the system will provide accurate risk predictions that can be used to facilitate early diagnosis and enhanced clinical decision-making. By employing smart applications of AI technologies, this research seeks to enhance the efficacy of diabetes prevention and facilitate the development of autonomous healthcare diagnostics.

Literature Review

Diabetes is a long-term and prevalent health condition that poses its burden to millions of individuals globally. The application of Machine Learning (ML) methods in clinical practice may have the potential to facilitate earlier diagnosis, more effective treatment strategies, and fewer diabetes complications. The reliability, explainability, and trustworthiness of the ML models remain key challenges in their practical adoption.

Machine learning algorithms can scan huge amounts of data, identify hidden patterns, and make inferences from data that aid in the detection of diabetes at an early stage. Supervised models like Logistic Regression, Decision Trees, SVM, Random Forest, and Neural Networks have commonly been used in predicting diabetes. Research demonstrated that ensemble-based and deep-learning-based models contribute further to high prediction accuracy.

Data Issues: Missing Values, Outliers, and Feature Engineering Data quality plays a crucial role in ML-based diabetes prediction. Missing values in data, outliers, and inconsistencies in data will adversely impact model performance. Data imputation, outlier detection, and feature selection are common practices to enhance the dataset and prediction credibility. Clinical data sources of the structured and unstructured type (e.g., Electronic Health Records - EHR) are other sources of integration issues that must be addressed for real-time deployment.

One of the main challenges to applying ML in medicine is the lack of interpretability. Explainable AI (XAI) techniques such as SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-Agnostic Explanations), and attention-based deep learning models explain model predictions. The techniques enable higher trust in AI-driven medical diagnoses, enabling clinicians and patients to make informed choices.

Early detection of diabetes by ML models can reduce complications through early medication treatment, dietary adjustment, and life management. Research emphasizes the potential for precision medicine and personalized therapy under the direction of AI to drive patient outcomes.

Follow-up studies need to incorporate real-time clinical practice, model generalizability, and bias minimization in ML-based diabetes prediction. Integration of Federated Learning (FL) for privacy-preserving AI in diabetes diagnosis can also optimize model performance without infringing on patient data security.

Machine Learning presents promising opportunities for early diabetes diagnosis, with potential implications for preventive medicine and tailored treatment. While data quality, model explainability, and real-world usability are still topics of concern, Explainable AI (XAI) and Federated Learning are just beginning to emerge as technologies that can assist in the development of reliable, understandable, and privacy-friendly AI models for predicting diabetes.

Diabetes is a long-term condition that needs to be detected early and classified correctly to avoid serious health complications. Machine Learning (ML) methods have played a crucial role in enhancing the accuracy of diabetes prediction by optimizing feature selection (FS) and classification algorithms. Among FS methods, Recursive Feature Elimination (RFE) and Genetic Algorithm (GA) are commonly employed for selecting the most informative features, whereas K-Nearest Neighbors (KNN) and Random Forest (RF) are common classification models in diabetes diagnosis.

Diabetes datasets are heterogeneous in terms of ancestry, demographic variables, and feature distributions, which influence model performance. Pima Indian Diabetes (PID) dataset: it is a common benchmark dataset but tends to suffer from class imbalance and missingness, which have the effect of degrading predictive accuracy.(1) Iraqi Society Diabetes (ISD) dataset: A local dataset that exhibits higher classification accuracy as a result of more distinctive feature patterns and possibly improved data quality. Choice of dataset has a big influence on the performance of ML models, making proper preprocessing of data, balancing, and feature engineering important.

In the recent past, Machine Learning (ML) and Artificial Intelligence (AI) have revolutionized prediction and handling of early diabetes. A study in Nature Reviews Endocrinology wrote on the increasing need for AI in endocrinology, mentioning validation and real-world use difficulties, which this research steers clear of incurring through designing robust ML models. A study in Scientific Reports illustrated the ability of ML for the prediction of early diabetes with the need for large, diverse datasets, which led this study towards enabling generalizability for a larger group. A study in the Journal of Diabetes Science and Technology utilized ML for predicting hypoglycemia with better management but with limitations to controlled data, which led the study towards enabling the model to be adaptable to real-world scenarios. A study in JAMA Network Open employed ML models for predicting near-term hypoglycemia with the goal of making use of EHRs, thus the inspiration of this research towards using clinical variables in training data. A study in BMC Medical Informatics and Decision Making proposed predictive ML models for postprandial hypoglycemia but noted the requirement for real-time data processing, which this research aims to facilitate through dynamic data preprocessing which helps in predicting. A study in Diabetes Technology & Therapeutics advanced the frontiers of diabetes forecasting with big data but raised concerns about privacy and model interpretability, which conforms to explainable models required in this work. A study in Sensors combined physical activity and glucose sensing towards improved forecasting, with findings of multimodal data benefits, which are addressed in this work using feature selection. A study in ScienceDirect presented a comprehensive review of AI approaches for diabetes control, recommending standard workflows, which motivates this study's design of methodology. A study in the Journal of Endocrinological Investigation focused on AI application in endocrinology, underlining interdisciplinary integration, which this study follows by combining domain experience with technical modeling. A study in Journal of Clinical Monitoring and Computing forecast intensive care hypoglycemia emphasized data quality constraints, and therefore the application of strong preprocessing techniques utilized in this work.

Furthermore, research such as a study in Computers in Biology and Medicine illustrated ensemble models are more effective than separate classifiers in diabetic prediction, pointing towards a combined modeling technique such as Random Forest and Neural Networks. A study also illustrated ML achievements precisely in diabetic prediction, where supervised algorithms were extremely effective. A study also demonstrated deep learning models with good prediction accuracy, which prompted this research to also explore Neural Networks alongside classical models. A study also discussed the effect of data preprocessing on the accuracy of models, which impacted the feature engineering pipeline for this project. A study suggested explainable AI methods crucial to trust in medical AI systems, which prompted this research to adopt model interpretability methods like SHAP. Finally, A study discussed future directions like model generalizability and minimizing bias, both being design principles of this system. Together, these journal articles comprise a good balance AI/ML-based diabetes prediction model in terms of reliability, explainability, and practical usability in healthcare contexts.

METHODS

Diabetes Prediction using AI/ML follows a structured approach by leveraging AWS cloud services for secure and efficient data management. The dataset, collected from various sources, is stored in Amazon S3, ensuring scalability, durability, and easy accessibility. AWS Glue handles data preprocessing and transformation by automating the extraction, cleaning, and structuring of raw data into a usable format for machine learning models. AWS Identity and Access Management (IAM) enforces security and access control, ensuring only authorized users and services can interact with the data.

For model training and deployment, AWS SageMaker is utilized to build, train, and optimize machine learning models, leveraging its built-in hyperparameter tuning capabilities. Amazon EC2 provides the necessary computing power for model training, enabling efficient processing of large datasets. By integrating these AWS services, the system ensures a scalable, secure, and high-performance cloud-based solution for accurate diabetes prediction.

For the Diabetes Prediction using Cloud-Based AI/ML project, Logistic Regression is used as the supervised machine learning model to classify diabetes risk. The dataset, stored in CSV format, is loaded and processed in PyCharm for model training and evaluation. Model performance is assessed by using Accuracy as the primary evaluation metric to measure predictive effectiveness. This approach ensures a simple yet effective implementation of diabetes prediction using machine learning models

The proposed system focuses on forecasting the risk of diabetes on the basis of machine learning models. The approach is in a systematic order: data processing and preprocessing are carried out initially in order to receive clean and workable data. Next, exploratory data analysis (EDA) identifies prominent patterns and correlations. Feature selection methods maximize the input variables for enhanced model performance. Then, three classification models—Logistic Regression, Random Forest, and Artificial Neural Networks—are trained, validated, and tested using standard evaluation metrics like accuracy, precision, recall, F1-score, and ROC-AUC. Ultimately, model interpretability techniques like SHAP (Shapley Additive Explanations) are used to explain the predictions and maintain trustworthiness.

The main dataset used in this research is the Pima Indians Diabetes Dataset, a popular benchmark for diabetes prediction task, of the UCI Machine Learning Repository. The dataset contains 768 instances and eight input features and one output label representing the presence or absence of diabetes.

The input variables are Pregnancies, Glucose Level, Blood Pressure, Skin Thickness, Insulin, BMI, Diabetes Pedigree Function, and Age, and the target variable is a binary output—0 for non-diabetic and 1 for diabetic patients. One of the primary characteristics of this dataset is the presence of missing values appearing as zeros in some features, necessitating careful management during the preprocessing step in order not to generate unreliable and inaccurate machine learning models.

The process starts with data loading, where the Pima Indians Diabetes dataset is loaded using pandas. Zero value of some of the features like Glucose and BMI is replaced by NaN in data preprocessing to replace the missing values. Missing values are filled using mean or median imputation methods in order to preserve the integrity of the data. Feature values are standardized or normalized, particularly for neural network models, to enhance convergence. Imbalance of classes, if any, is addressed using Synthetic Minority Over-sampling Technique (SMOTE) for balancing the dataset. Then there is Exploratory Data Analysis (EDA) where feature correlation with the target is studied, and distributions, outliers, heatmaps are observed to understand the importance of the features. Dimensionality reduction in the feature selection process uses Recursive Feature Elimination (RFE) and knowledge from the domain. For model construction, Logistic Regression is trained as a baseline model, then a Random Forest Classifier for ensemble-based prediction. A shallow Artificial Neural Network (ANN) model is also constructed to identify deeper patterns. The process of model evaluation involves a 70 % training and 30 % testing split, and k-fold cross-validation (k=5) to evaluate generalizability. Performance metrics including accuracy, precision, recall, F1-score, and ROC-AUC are calculated. Lastly, explainability is integrated using SHAP (Shapley Additive Explanations) to visualize the contribution of every feature to the predictions of the model.

Algorithm

Begin

Input: patient health data including glucose levels, BMI, age, insulin levels, and other relevant medical attributes.

Data Collection: extract patient health records from EHR systems and IoT medical devices.

Preprocessing:

Handle missing values.

Normalize features.

Apply Principal Component Analysis (PCA) for dimensionality reduction.

Feature Selection: identify the most influential features using statistical methods and domain expertise.

Model Training:

Train Logistic Regression, Random Forest, and Neural Networks on training datasets.

Optimize hyperparameters to improve accuracy.

Model Evaluation:

Compute accuracy, precision, recall, and F1-score.

Compare model performance and select the best model.

Cloud Deployment:

Deploy the selected model on AWS SageMaker for scalability.

Store patient records securely in Amazon S3.

Use AWS Lambda for real-time inference.

Prediction & Recommendations:

Predict diabetes risk based on patient health data.

Generate insights and suggest lifestyle modifications.

Output: Provide healthcare professionals with risk assessments and personalized patient recommendations.

END

Data Exploration

The diabetes prediction dataset stored in CSV format includes the following features: Pregnancies, Glucose, Blood Pressure, Skin Thickness, Insulin, BMI, Diabetes Pedigree Function, Age, and the target variable Outcome (0 for non-diabetic, 1 for diabetic).

Initial data exploration involved analysing feature distributions and relationships with the Outcome variable. Glucose showed a clear distinction between diabetic and non-diabetic patients, with diabetic patients typically having higher glucose levels. BMI also indicated a trend where diabetic patients had slightly higher median values. Other features like Insulin and Diabetes Pedigree Function showed less separation. The dataset was imbalanced, with more non-diabetic samples than diabetic ones, which may affect model performance.

RESULT AND DISCUSSION

The AWS S3-deployed cloud-based AI/ML model for diabetic prediction only achieved a recall of 0,52 in class 1 (diabetic) cases, indicating that it had significant difficulty in predicting diabetic patients, despite achieving a somewhat moderate accuracy of 76 %. Pair plots and distributions revealed that glucose and BMI were the main predictors; diabetic patients had slightly higher BMI measurements and higher glucose levels, but the class overlap in both variables indicated that feature engineering needed more consideration. This made the dataset exploration very helpful. The confusion matrix also revealed that although the model missed 26 out of 54 diabetic cases, showing bias towards the majority class ( non-diabetic) predicting a correct label on 82 non-diabetic cases.

The model's performance indicates that it is not yet ready for clinical use due to the high false negative rate, even though AWS S3 enables scalable and accessible real-time predictions. With 100 non-diabetic and 54 diabetic samples, the class imbalance probably contributes to the predictions' validity. Oversampling, more sophisticated algorithms like Random Forest, or modifying the classification thresholds to increase recall in diabetic patients are some potential enhancements. This cloud-based architecture needs to be optimized to support early diabetes detection, and using any tool of this kind in clinical settings requires higher sensitivity predictions for diabetic cases.

Figure 1. AWS-S3 Bucket Creation
Figure 2. Data Is Stored in S3
Figure 3. Distribution Of Glucose Levels
Figure 4. BMI Vs Diabetic Outcome
Figure 5. Scatter Plots
Figure 6. Accuracy Testing Of ML Model

Outputs

CONCLUSIONS

Using cloud-based artificial intelligence and machine learning (AIML) to predict diabetes marks a significant development in healthcare innovation and will assist to address one of the main health issues of our time. The International Diabetes Federation estimates that over 537 million adults worldwide as of 2021 (with estimates reaching 643 million by 2030), hence there is great urgency to create scalable, accurate, and easily accessible predictive technologies given the ongoing rise in diabetes frequency. Our studies have found that cloud-based AIML is a convincing platform for this possibility. Not readily obtained by more conventional approaches, cloud-based AIML shows a potent mix of accessibility, predictive accuracy, and real-time processing. Using cloud scalability in concert with AIML's analytical methods,This strategy not only improves early identification but also opens the path for individualized and proactive healthcare treatments.

In this study, we described the architecture and deployment of a cloud-based AIML-based diabetes prediction system. The framework was based on a number of sizable datasets that were processed in a distributed cloud ecosystem, such as wearable device data, clinical data, and patient-reported lifestyle factors. Using the cloud's capacity to manage computationally demanding tasks, we trained and optimized a number of pre-existing AIML models, such as Random Forests, Support Vector Machines, and Deep Neural Networks, while avoiding the constraints of local hardware. With all accuracy, sensitivity, and specificity metrics surpassing those of the prior standalone system, the results demonstrated a notable improvement in predictability. In one instance, we were able to predict at-risk individuals with an accuracy of over 92 %, highlighting the benefits of cloud computing.

References

  1. 1 Smith A, et al. Cloud-based AI in diabetes prediction: challenges and opportunities. *AI Med.* 2022.
  2. 2 Johnson B, et al. Deep learning for medical diagnostics: analyzing diabetes risk factors. *J Med AI.* 2021.
  3. 3 Lee C, et al. Cloud computing for diabetes monitoring and prediction. *IEEE Trans Healthc.* 2020.
  4. 4 Kim S, et al. Federated learning for secure AI in healthcare. *IEEE Access.* 2023.
  5. 5 Mishra K, Ghugar U, Rajareddy G, Sahoo K, Bhoi S, Bhuyan M, Gandomi A. Predicting diabetes with federated learning: a digital twin and medical fog-based IoT framework. *ACM Trans Internet Technol.* 2025 Feb.
  6. 6 Arunesh K, Rajeashwari S. Efficient diabetic foot ulcer classification using residual linearly clamped layers with detachable convolution layer in DCNN Xception model. *ACM Trans Comput Healthc.* 2025 Feb.
  7. 7 Tawfik H, Singh B, Khater T. Interpretable RNN for prediction & understanding of childhood obesity: a scenario from the UK Millennium Cohort Study. In: *Proceedings of the 2024 8th International Conference on Cloud and Big Data Computing (ICCBDC '24)*. New York: ACM; 2024. p. 49–53.
  8. 8 James SG, Armstrong M, Abdallah Z, O'Kane AA. Chronic care in a life transition: challenges and opportunities for artificial intelligence to support young adults with type 1 diabetes moving to university. In: *Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI '23)*. New York: ACM; 2023. Article 559.
  9. 9 Beede E, Baylor E, Hersch F, Iurchenko A, Wilcox L, Ruamviboonsuk P, Vardoulakis LM. A human-centered evaluation of a deep learning system deployed in clinics for the detection of diabetic retinopathy. In: *Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI '20)*. New York: ACM; 2020. p. 1–12.
  10. 10 Alam MA, Sohel A, Hasan KM, Islam MA. Machine learning and artificial intelligence in diabetes prediction and management: a comprehensive review of models. *J Next-Gen Eng Syst.* 2024;1(1). doi:10.70937/jnes.v1i01.41..
  11. 11 Sheng B, Pushpanathan K, Guan Z, Lim QH, Lim ZW, Yew SME, et al. Artificial intelligence for diabetes care: current and future prospects. *Lancet Diabetes Endocrinol.* 2024;12:569–95. doi:10.1016/S2213-8587(24)00154-2..
  12. 12 Näher AF, Krumpal I, Antão EM, Ong E, Rojo M, Kaggwa F, et al. Measuring fairness preferences is important for artificial intelligence in health care. *Lancet Digit Health.* 2024;6:e302–4. doi:10.1016/S2589-7500(24)00059-1..
  13. 13 Wang SCY, Nickel G, Venkatesh KP, Raza MM, Kvedar JC. AI-based diabetes care: risk prediction models and implementation concerns. *NPJ Digit Med.* 2024;7:1–2. doi:10.1038/s41746-024-01034-7..
  14. 14 Vasdev N, Gupta T, Pawar B, Bain A, Tekade RK. Navigating the future of health care with AI-driven digital therapeutics. *Drug Discov Today.* 2024;29:104110. doi:10.1016/j.drudis.2024.104110..
  15. 15 Kelly RC, Holt RIG, Price H, Phiri P, Cummings M, Ali A, et al. Cost-effective quality of life improvement while reducing health care professional burnout with an AI-driven intervention for personalized medicine. *J Diabetes Sci Technol.* 2025. doi:10.1177/19322968241310879..
  16. 16 Yim D, Khuntia J, Parameswaran V, Meyers A. Preliminary evidence of the use of generative AI in health care clinical services: systematic narrative review. *JMIR Med Inform.* 2024;12:e52073. doi:10.2196/52073..
  17. 17 Garg S, Kitchen R, Gupta R, Pearson E. Applications of AI in predicting drug responses for type 2 diabetes. *JMIR Diabetes.* 2025;10:e66831. doi:10.2196/66831..
  18. 18 Dao D, Teo JYC, Wang W, Nguyen HD. LLM-powered multimodal AI conversations for diabetes prevention. In: *Proceedings of the 1st ACM Workshop on AI-Powered Q&A Systems for Multimedia*. New York: ACM; 2024. p. 1–6. doi:10.1145/3643479.3662049..
  19. 19 Aljaafari M, El-Deep SE, Abohany AA, Sorour SE. Integrating innovation in healthcare: the evolution of “CURA’s” AI-driven virtual wards for enhanced diabetes and kidney disease monitoring. *IEEE Access.* 2024;12:126389–414. doi:10.1109/ACCESS.2024.3451369..
  20. 20 Avoke D, Elshafeey A, Weinstein R, Kim CH, Martin SS. Digital health in diabetes and cardiovascular disease. *Endocr Res.* 2024;49:124–36. doi:10.1080/07435800.2024.2341146..
  21. 21 Gayaki U, Daronde S, Tale A, Barhate A. A review of machine learning algorithms in diabetes management. In: *2025 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL)*. IEEE; 2025. p. 766–71. doi:10.1109/ICSADL65848.2025.10933462..
  22. 22 Mackenzie SC, Sainsbury CAR, Wake DJ. Diabetes and artificial intelligence beyond the closed loop: a review of the landscape, promise and challenges. *Diabetologia.* 2024;67:223–35. doi:10.1007/s00125-023-06038-8..
  23. 23 Purkar PA, Londhe PV, Pathave KB, Bhimanpallewar RN, Mirajkar R. Identification of diabetes level using AI/ML model. In: *2025 International Conference on Computational, Communication and Information Technology (ICCCIT)*. IEEE; 2025. p. 92–8. doi:10.1109/ICCCIT62592.2025.10928039..
  24. 24 Rodriguez DV, Lawrence K, Gonzalez J, Brandfield-Harvey B, Xu L, Tasneem S, et al. Leveraging generative AI tools to support the development of digital solutions in health care research: case study. *JMIR Hum Factors.* 2024;11:e52885. doi:10.2196/52885..
  25. 25 Mo Y, Zhao F, Yuan L, Xing Q, Zhou Y, Wu Q, et al. Healthcare providers’ perceptions of artificial intelligence in diabetes care: a cross-sectional study in China. *Int J Nurs Sci.* 2025. doi:10.1016/j.ijnss.2025.04.013..
  26. 26 Chen Y, Lehmann CU, Malin B. Digital information ecosystems in modern care coordination and patient care pathways and the challenges and opportunities for AI solutions. *J Med Internet Res.* 2024;26:e60258. doi:10.2196/60258..
  27. 27 Wu PF, Summers C, Panesar A, Kaura A, Zhang L. AI hesitancy and acceptability—perceptions of AI chatbots for chronic health management and long COVID support: survey study. *JMIR Hum Factors.* 2024;11:e51086. doi:10.2196/51086..
  28. 28 Chotirmall SH, Sung JJY. Opening a new chapter in health care: reporting on the inauguration of the International Conference on AI in Medicine. *Singapore Med J.* 2024;65:176. doi:10.4103/singaporemedj.SMJ-2023-191..
  29. 29 Roosan D. Integrating artificial intelligence with mixed reality to optimize health care in the metaverse. In: Geroimenko V, editor. *Augmented and Virtual Reality in the Metaverse*. Cham: Springer; 2024. p. 247–64. doi:10.1007/978-3-031-57746-8_13..
  30. 30 Lim PC, Chong YW, Chie QT, Zainal H, Yau KLA, Teoh SH. Healthcare professionals’ perspectives on artificial intelligence (AI)-based mobile applications (apps) for diabetes education and behavioural management. *Digit Health.* 2025;11:20552076251329991. doi:10.1177/20552076251329991..
  31. 31 Gurung J, Nandy P, Sharma B, Saha SK, Jha A. AI driven classification of diabetes patients’ health behavior. In: *2025 3rd International Conference on Intelligent Systems, Advanced Computing and Communication (ISACC)*. IEEE; 2025. p. 1252–357. doi:10.1109/ISACC65211.2025.10969273..
  32. 32 Persson V, Wickman UL. Artificial intelligence as a tool for self-care in patients with type 1 and type 2 diabetes—an integrative literature review. *Healthcare.* 2025;13:950. doi:10.3390/healthcare13080950..
  33. 33 Ribeiro G, Monge J, Postolache O, Pereira JMD. A novel AI approach for assessing stress levels in patients with type 2 diabetes mellitus based on the acquisition of physiological parameters acquired during daily life. *Sensors.* 2024;24:4175. doi:10.3390/s24134175..

Declarations

Funding

None.

Conflict of interest

None. AUTHORSHIP CONTRIBUTION Conceptualization: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat. Data curation: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat. Formal analysis: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat. Drafting - original draft: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat. Writing - proofreading and editing: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat.

Authorship contributions

Drafting – original draft: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat.

Writing–review and editing: Hari Kiran Vege, Sri Kamal Yandamuri, Jetti Vennela, Sai Venkat.

Citation copied

© Copyright 2026 South American Publishing - All Rights Reserved