Salvatore Corrao
Miodrag Janić
Viviana Maggio
Manfredi Rizzo- 1Department of Clinical Medicine, Internal Medicine Unit, National Relevance and High Specialization Hospital Trust Azienda di Rilievo Nazionale ed Alta Specializzazione (ARNAS) Civico Di Cristina Benfratelli, Palermo, Italy
- 2School of Medicine, Promise Department of Health Promotion Sciences Maternal and Infantile Care, Internal Medicine and Medical Specialties, University of Palermo, Palermo, Italy
- 3Department of Endocrinology, Diabetes and Metabolic Diseases, University Medical Centre Ljubljana, Ljubljana, Slovenia
- 4Faculty of Medicine, University of Ljubljana, Ljubljana, Slovenia
- 5Internal Medicine Department, Ras Al Khaimah College of Medical Sciences, Ras Al Khaimah Medical and Health Sciences University, Ras Al Khaimah, United Arab Emirates
1 Introduction
Diabetes is a chronic disease affecting over 400 million people globally, with its prevalence expected to rise sharply in the coming decades due to ageing populations, sedentary lifestyles, and increasing obesity rates (1). It is a leading cause of serious complications, including cardiovascular disease, nephropathy, neuropathy, and retinopathy, contributing significantly to morbidity, mortality, and healthcare costs (2, 3). Traditional approaches to diabetes management often fail to address the complexity of the disease, particularly in cases involving highly variable glucose patterns or multiple comorbidities (4–6).
The emergence of artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), has introduced new possibilities in diabetology. While ML focuses on analyzing structured datasets to identify patterns and trends, DL extends this capability by processing unstructured and multimodal data, such as retinal images and continuous glucose monitoring (CGM) outputs (7, 8). These technologies create a paradigm shift in diabetes care, empowering clinicians and patients with precise, data-driven insights.
This opinion focuses on the applications, methodological differences, and critical advancements in using ML and DL in diabetes. With the exponential growth of healthcare data generated from wearable devices, CGMs, and electronic health records (EHRs), there is immense potential to enhance diabetes management. However, the complexity of this chronic disease demands more sophisticated tools than traditional methods can offer.
2 Differences between machine learning and deep learning
ML and DL are subfields of AI that share a common goal of enabling machines to learn from data, yet they differ fundamentally in their methodologies, complexity, and applications. ML focuses on algorithms that analyze structured data and make predictions or decisions based on these patterns. Standard techniques include logistic regression, decision trees, support vector machines (SVMs), and random forests. These models typically rely on manual feature extraction, where domain experts determine the most relevant variables for analysis (9, 10).
In contrast, DL uses artificial neural networks inspired by the human brain to process data hierarchically. It excels in analyzing unstructured data, such as images, text, and sequential data, without explicit feature engineering. Architectures like convolutional neural networks (CNNs) are particularly effective for image analysis. At the same time, recurrent neural networks (RNNs) and long short-term memory (LSTM) networks are widely applied to time-series data (11, 12). DL models, however, demand significantly larger datasets and greater computational power than traditional ML approaches, making them more resource-intensive (13).
The choice between ML and DL depends mainly on the nature of the problem, data type, and computational resources. ML is often preferred for tasks involving smaller datasets and requiring interpretability, such as predicting diabetes risk based on clinical and demographic data. DL, by contrast, is better suited for high-dimensional data analysis, such as detecting diabetic retinopathy from retinal fundus images or forecasting glucose fluctuations from CGM readings (14, 15). These differences underscore the complementary roles of ML and DL in advancing diabetes care as depicted in Figure 1.
Figure 1. Applications of machine learning and deep learning in the practice of diabetology (see the text for more explanation).
3 Applications of machine learning in diabetology
3.1 Early detection and risk prediction
ML algorithms are pivotal in identifying individuals at risk of developing type 2 diabetes or its complications. Logistic regression and random forest models, for example, analyze demographic and clinical factors, such as body mass index (BMI), fasting glucose, and family history, to calculate risk scores (16, 17). Moreover, these models have been employed to stratify patients based on their likelihood of developing complications like diabetic retinopathy or nephropathy, facilitating timely medical interventions (18).
3.2 Personalized treatment
Personalized medicine is another transformative application of ML in diabetology. To optimize treatment regimens, algorithms analyze patient-specific data, such as glucose patterns, activity levels, and dietary intake. Ensemble learning models have successfully predicted insulin dosages tailored to individual needs, reducing the risks of hypoglycemia and hyperglycemia (19, 20).
3.3 Resource optimization
ML also addresses resource challenges in healthcare systems by predicting hospitalization risks and prioritizing high-risk patients for follow-up care. EHR-based predictive models enable clinicians to allocate resources effectively, ensuring timely interventions for those with poorly controlled diabetes (21).
4 Applications of deep learning in diabetology
4.1 Continuous glucose monitoring
DL has significantly enhanced CGM systems by improving the accuracy of glucose prediction models. RNNs and LSTM networks analyze sequential glucose data to forecast levels several hours ahead, enabling patients to adjust insulin or carbohydrate intake proactively (22–24).
4.2 Automated insulin delivery systems
DL algorithms power artificial pancreas systems by integrating CGM data with insulin pump controls. CNNs detect patterns in glucose trends, automating insulin delivery with precision. Clinical trials have shown that these systems improve glycemic control and reduce HbA1c levels (25).
4.3 Retinopathy detection
CNNs are widely used in analyzing retinal fundus images for the early detection of diabetic retinopathy. These models have achieved diagnostic accuracies comparable to human ophthalmologists, making them particularly valuable in resource-constrained settings (12, 26).
4.4 Cardiovascular and neuropathy risk prediction
DL models also help assess the risks of diabetic complications, such as cardiovascular diseases and neuropathy. Transformer-based architectures, such as BERT, combine data from wearable devices, clinical notes, and EHRs to provide comprehensive risk assessments (17).
5 Challenges in implementing machine learning and deep learning in diabetology
Despite their transformative potential, adopting ML and DL in clinical settings faces several challenges. Data quality and standardization remain significant barriers, as healthcare datasets often need to include more consistent values (10). Moreover, the “black box” nature of DL models limits their interpretability, making it difficult for clinicians to trust and adopt these tools (27). Ethical issues, including data privacy and biases in training datasets, further complicate their implementation (28, 29).
Emerging solutions like federated learning allow decentralized model training without compromising patient privacy. Explainable AI (XAI) initiatives are improving transparency fostering greater trust among healthcare professionals (30).
6 Discussion
This opinion provides actionable insights by discussing AI’s opportunities in diabetology and critically analyzing its current limitations. Researchers, clinicians, and healthcare professionals should be aware how these transformative technologies can significantly improve the lives of people with diabetes as ML and DL revolutionize diabetes management by enabling early detection, personalized treatment, and advanced monitoring.
Recent advances in artificial intelligence have extended beyond individualized care to offer strategic insights into diabetes care trends at the population level. AI tools, particularly interpretable machine learning models such as the Logic Learning Machine, have been successfully applied to identify predictors of lipid goal attainment in large outpatient cohorts with type 2 diabetes, as demonstrated by the AMD Artificial Intelligence Study Group (31). Similarly, deep learning applied to electronic health records has enabled the prognostic modelling of incident heart failure among diabetic patients, offering an early warning system for adverse cardiovascular outcomes (32). In another large-scale application, transparent ML algorithms have revealed patterns of therapeutic inertia in HbA1c trajectories, shedding light on modifiable gaps in diabetes management at the system level (33). These findings underscore the expanding scope of AI in addressing not only personalized medicine, but also healthcare planning, quality assessment, and intervention targeting on a broader scale.
The future of ML and DL in diabetology is promising. While challenges related to data quality, interpretability, and ethics persist, advances in wearable technology and real-time data integration will further enhance monitoring capabilities, overcoming these barriers. Federated learning frameworks will facilitate collaboration between institutions, while XAI will make AI tools more interpretable. Continued investment in AI infrastructure and interdisciplinary research will be critical to realize the full potential of these technologies to transform diabetes care. Integrating AI into clinical workflows will undoubtedly improve outcomes and enhance the quality of care for millions of individuals living with diabetes.
Author contributions
SC: Conceptualization, Writing – original draft, Writing – review & editing. MJ: Writing – review & editing. VM: Writing – review & editing. MR: Writing – review & editing.
Funding
The author(s) declare that no financial support was received for the research and/or publication of this article.
Conflict of interest
The authors declare that the manuscript was prepared in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The author(s) declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.
Generative AI statement
The author(s) declare that no Generative AI was used in the creation of this manuscript.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1. International Diabetes Federation. IDF Diabetes Atlas. 10th ed. Brussels, Belgium: International Diabetes Federation (2021). Available at: https://diabetesatlas.org.
2. Fowler MJ. Microvascular and macrovascular complications of diabetes. Clin. Diabetes. (2008) 26:77–82. doi: 10.2337/diaclin.26.2.77
3. Diabetes C, Complications Trial Research G, Nathan DM, Genuth S, Lachin J, Cleary P, et al. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl. J. Med. (1993) 329:977–86. doi: 10.1056/NEJM199309303291401
4. Forlenza GP. Use of artificial intelligence to improve diabetes outcomes in patients using multiple daily injections therapy. Diabetes Technol. Ther. (2019) 21:S24–S8. doi: 10.1089/dia.2019.0077
5. Herder C, Rizzo M, and Roden M. Precision diabetology: Where do we stand now? J. Diabetes Complications. (2024) 38:108899. doi: 10.1016/j.jdiacomp.2024.108899
6. Rosta L, Menyhart A, Mahmeed WA, Al-Rasadi K, Al-Alawi K, Banach M, et al. Telemedicine for diabetes management during COVID-19: what we have learnt, what and how to implement. Front. Endocrinol. (Lausanne). (2023) 14:1129793. doi: 10.3389/fendo.2023.1129793
7. Shao J, Pan Y, Kou WB, Feng H, Zhao Y, Zhou K, et al. Generalization of a deep learning model for continuous glucose monitoring-based hypoglycemia prediction: algorithm development and validation study. JMIR Med. Inform. (2024) 12:e56909. doi: 10.2196/56909
8. Gulshan V, Peng L, Coram M, Stumpe MC, Wu D, Narayanaswamy A, et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA. (2016) 316:2402–10. doi: 10.1001/jama.2016.17216
9. Biester T, Tauschmann M, Chobot A, Kordonouri O, Danne T, Kapellen T, et al. The automated pancreas: A review of technologies and clinical practice. Diabetes Obes. Metab. (2022) 24 Suppl 1:43–57. doi: 10.1111/dom.14576
10. Torkamani A, Andersen KG, Steinhubl SR, and Topol EJ. High-definition medicine. Cell. (2017) 170:828–43. doi: 10.1016/j.cell.2017.08.007
11. Miotto R, Li L, Kidd BA, and Dudley JT. Deep patient: an unsupervised representation to predict the future of patients from the electronic health records. Sci. Rep. (2016) 6:26094. doi: 10.1038/srep26094
12. Guan Z, Li H, Liu R, Cai C, Liu Y, Li J, et al. Artificial intelligence in diabetes management: Advancements, opportunities, and challenges. Cell Rep. Med. (2023) 4:101213. doi: 10.1016/j.xcrm.2023.101213
13. Lipton ZC. The mythos of model interpretability. Commun. ACM. (2018) 61:36–43. doi: 10.1145/3233231
14. McMahan B, Moore E, Ramage D, Hampson S, and Arcas B. Communication-efficient learning of deep networks from decentralized data. In: Aarti S and Jerry Z, editors. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics. Proceedings of Machine Learning Research: PMLR (2017). p. 1273–82.
15. Doshi-Velez F and Kim B. Towards A rigorous science of interpretable machine learning. arXiv: Mach. Learn. (2017). doi: 10.48550/arXiv.1702.08608
16. Esteva A, Robicquet A, Ramsundar B, Kuleshov V, DePristo M, Chou K, et al. A guide to deep learning in healthcare. Nat. Med. (2019) 25:24–9. doi: 10.1038/s41591-018-0316-z
17. Oikonomou EK and Khera R. Machine learning in precision diabetes care and cardiovascular risk prediction. Cardiovasc. Diabetol. (2023) 22:259. doi: 10.1186/s12933-023-01985-3
18. Obermeyer Z and Emanuel EJ. Predicting the future - big data, machine learning, and clinical medicine. N Engl. J. Med. (2016) 375:1216–9. doi: 10.1056/NEJMp160618
19. Clifton L, Clifton DA, Pimentel MA, Watkinson PJ, and Tarassenko L. Predictive monitoring of mobile patients by combining clinical observations with data from wearable sensors. IEEE J. BioMed. Health Inform. (2014) 18:722–30. doi: 10.1109/JBHI.2013.2293059
20. Chen JH and Asch SM. Machine learning and prediction in medicine - beyond the peak of inflated expectations. N Engl. J. Med. (2017) 376:2507–9. doi: 10.1056/NEJMp1702071
22. Vekic J, Silva-Nunes J, and Rizzo M. Glucose metabolism disorders: challenges and opportunities for diagnosis and treatment. Metabolites. (2022) 12(8):712. doi: 10.3390/metabo12080712
23. LeCun Y, Bengio Y, and Hinton G. Deep learning. Nature. (2015) 521:436–44. doi: 10.1038/nature14539
24. Schmidhuber J. Deep learning in neural networks: an overview. Neural Netw. (2015) 61:85–117. doi: 10.1016/j.neunet.2014.09.003
26. Organizing Committee of the Madrid Critical Care D, Nunez Reiz A, Martinez Sagasti F, Alvarez Gonzalez M, Blesa Malpica A, Martin Benitez JC, et al. Big data and machine learning in critical care: Opportunities for collaborative research. Med. Intensiva (Engl Ed). (2019) 43:52–7. doi: 10.1016/j.medine.2018.06.006
27. Jordan MI and Mitchell TM. Machine learning: Trends, perspectives, and prospects. Science. (2015) 349:255–60. doi: 10.1126/science.aaa8415
28. Elendu C, Amaechi DC, Elendu TC, Jingwa KA, Okoye OK, John Okah M, et al. Ethical implications of AI and robotics in healthcare: A review. Med. (Baltimore). (2023) 102:e36671. doi: 10.1097/MD.0000000000036671
29. Norori N, Hu Q, Aellen FM, Faraci FD, and Tzovara A. Addressing bias in big data and AI for health care: A call for open science. Patterns (NY). (2021) 2:100347. doi: 10.1016/j.patter.2021.100347
30. Muhammad D and Bendechache M. Unveiling the black box: A systematic review of Explainable Artificial Intelligence in medical image analysis. Comput. Struct. Biotechnol. J. (2024) 24:542–60. doi: 10.1016/j.csbj.2024.08.005
31. Masi D, Zilich R, Candido R, Giancaterini A, Guaita G, Muselli M, et al. Uncovering predictors of lipid goal attainment in type 2 diabetes outpatients using logic learning machine: insights from the AMD annals and AMD artificial intelligence study group. J. Clin. Med. (2023) 12(12):4095. doi: 10.3390/jcm12124095
32. Gandin I, Saccani S, Coser A, Scagnetto A, Cappelletto C, Candido R, et al. Deep-learning-based prognostic modeling for incident heart failure in patients with diabetes using electronic health records: A retrospective cohort study. PloS One. (2023) 18:e0281878. doi: 10.1371/journal.pone.0281878
33. Musacchio N, Zilich R, Masi D, Baccetti F, Nreu B, Bruno Giorda C, et al. A transparent machine learning algorithm uncovers HbA1c patterns associated with therapeutic inertia in patients with type 2 diabetes and failure of metformin monotherapy. Int. J. Med. Inform. (2024) 190:105550. doi: 10.1016/j.ijmedinf.2024.105550
Keywords: machine learning, deep learning, artificial intelligence, diabetes management, challenges
Citation: Corrao S, Janić M, Maggio V and Rizzo M (2025) Machine learning and deep learning in diabetology: revolutionizing diabetes care. Front. Clin. Diabetes Healthc. 6:1547689. doi: 10.3389/fcdhc.2025.1547689
Received: 18 December 2024; Accepted: 05 May 2025;
Published: 27 May 2025.
Edited by:
Michele Schiavon, University of Padua, ItalyReviewed by:
Ivan Pristas, Croatian Institute of Public Health, CroatiaCopyright © 2025 Corrao, Janić, Maggio and Rizzo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Miodrag Janić, bWlvZHJhZy5qYW5pY0BrY2xqLnNp