Advancing Diabetic Foot Ulcer Care: AI and Generative AI Approaches for Classification, Prediction, Segmentation, and Detection
Abstract
:1. Introduction
2. Methods
2.1. Research Questions
2.2. Inclusion and Exclusion Criteria
- Articles focused on the application of AI or generative AI in DFU;
- Article published during the period from 2020 to 2025;
- Studies in the English language.
- Articles focused on non-DFU-related medical imaging or conditions;
- Articles unrelated to AI applications in DFUs.
2.3. Search Strategy
2.4. Study Selection
3. Results
3.1. Classification
3.2. Prediction
3.3. Segmentation
3.4. Detection
3.5. Generative AI in DFU
3.6. Smartphone Applications for DFU
3.7. Evaluation Metrics and Datasets
4. Discussion
4.1. Roles of AI in DFU Management
4.2. Challenge
4.3. Future Direction
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Abbreviations
AI | artificial intelligence |
AFS | abrasion foot sores |
AUC | area under the receiver operating characteristic curve |
ANN | artificial neural network |
BPNN | backpropagation neural network |
BIM | body mass index |
CNN | convolutional neural network |
DM | diabetes mellitus |
DF | diabetic foot |
DFU | diabetic foot ulcer |
DFS | diabetic foot sores |
DFW | diabetic foot wounds |
DT | decision tree |
DICE | dice similarity coefficient |
DFUC | diabetic foot ulcer challenge |
EHR | electronic health record |
EMR | electronic medical record |
ELM | extreme learning machine |
FCL | fully connected layer |
FID | Fréchet inception distance |
GA | genetic algorithm |
GAN | generative adversarial network |
Grad-CAM | gradient-weighted class activation mapping |
HbA1C | hemoglobin 1C |
IoU | intersection over union |
KNN | k-nearest neighbor |
KID | kernel inception distance |
LR | logistic regression |
LIME | local interpretable model-agnostic explanations |
ML | machine learning |
mAP | mean average precision |
MAE | mean absolute error |
MLP | multilayer perceptron |
NB | naive bayes |
PVD | peripheral vascular disease |
PAD | peripheral artery disease |
RF | random forest |
RL | reinforcement learning |
ROC | receiver operating characteristic |
ReLU | rectified linear unit |
RAE | relative absolute error |
RMSE | root mean squared error |
SNN | Siamese neural network |
SVM | support vector machine |
SHAP | shapley additive explanations |
T2DM | type 2 diabetes mellitus |
TcPO₂ | transcutaneous oxygen pressure |
XAI | explainable artificial intelligence |
References
- Gosak, L.; Svensek, A.; Lorber, M.; Stiglic, G. Artificial Intelligence based prediction of diabetic foot risk in patients with diabetes: A literature review. Appl. Sci. 2023, 13, 2823. [Google Scholar] [CrossRef]
- Sun, H.; Saeedi, P.; Karuranga, S.; Pinkepank, M.; Ogurtsova, K.; Duncan, B.B.; Stein, C.; Basit, A.; Chan, J.C.; Mbanya, J.C. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res. Clin. Pract. 2022, 183, 109119. [Google Scholar] [CrossRef]
- Everett, E.; Mathioudakis, N. Update on management of diabetic foot ulcers. Ann. N. Y. Acad. Sci. 2018, 1411, 153–165. [Google Scholar] [CrossRef]
- Liu, C.; van Netten, J.J.; Van Baal, J.G.; Bus, S.A.; van Der Heijden, F. Automatic detection of diabetic foot complications with infrared thermography by asymmetric analysis. J. Biomed. Opt. 2015, 20, 026003. [Google Scholar] [CrossRef] [PubMed]
- Ndosi, M.; Wright-Hughes, A.; Brown, S.; Backhouse, M.; Lipsky, B.A.; Bhogal, M.; Reynolds, C.; Vowden, P.; Jude, E.; Nixon, J. Prognosis of the infected diabetic foot ulcer: A 12-month prospective observational study. Diabet. Med. 2018, 35, 78–88. [Google Scholar] [CrossRef] [PubMed]
- Kim, R.B.; Gryak, J.; Mishra, A.; Cui, C.; Soroushmehr, S.R.; Najarian, K.; Wrobel, J.S. Utilization of smartphone and tablet camera photographs to predict healing of diabetes-related foot ulcers. Comput. Biol. Med. 2020, 126, 104042. [Google Scholar] [CrossRef] [PubMed]
- Swaminathan, N.; Awuah, W.A.; Bharadwaj, H.R.; Roy, S.; Ferreira, T.; Adebusoye, F.T.; Ismail, I.F.N.b.; Azeem, S.; Abdul-Rahman, T.; Papadakis, M. Early intervention and care for Diabetic Foot Ulcers in Low and Middle Income Countries: Addressing challenges and exploring future strategies: A narrative review. Health Sci. Rep. 2024, 7, e2075. [Google Scholar] [CrossRef]
- Briganti, G.; Le Moine, O. Artificial intelligence in medicine: Today and tomorrow. Front. Med. 2020, 7, 509744. [Google Scholar] [CrossRef]
- Zhang, P.; Kamel Boulos, M.N. Generative AI in medicine and healthcare: Promises, opportunities and challenges. Future Internet 2023, 15, 286. [Google Scholar] [CrossRef]
- McCarthy, J. What Is Artificial Intelligence. 2007. Available online: http://jmc.stanford.edu/artificial-intelligence/what-is-ai/ (accessed on 5 October 2024).
- Pulapaka, S.; Godavarthi, S.; Ding, D.S. Introduction to Generative AI. In Empowering the Public Sector with Generative AI: From Strategy and Design to Real-World Applications; Springer: Berlin/Heidelberg, Germany, 2024; pp. 1–29. [Google Scholar]
- Guan, H.; Wang, Y.; Niu, P.; Zhang, Y.; Zhang, Y.; Miao, R.; Fang, X.; Yin, R.; Zhao, S.; Liu, J. The role of machine learning in advancing diabetic foot: A review. Front. Endocrinol. 2024, 15, 1325434. [Google Scholar] [CrossRef]
- Alzubaidi, L.; Fadhel, M.A.; Oleiwi, S.R.; Al-Shamma, O.; Zhang, J. DFU_QUTNet: Diabetic foot ulcer classification using novel deep convolutional neural network. Multimed. Tools Appl. 2020, 79, 15655–15677. [Google Scholar] [CrossRef]
- Liu, Z.; John, J.; Agu, E. Diabetic foot ulcer ischemia and infection classification using efficientnet deep learning models. IEEE Open J. Eng. Med. Biol. 2022, 3, 189–201. [Google Scholar] [CrossRef] [PubMed]
- Ahsan, M.; Naz, S.; Ahmad, R.; Ehsan, H.; Sikandar, A. A deep learning approach for diabetic foot ulcer classification and recognition. Information 2023, 14, 36. [Google Scholar] [CrossRef]
- Liu, Z.; Han, S.; Gao, L.; Wang, J.; Yao, Q. A Meaningful Machine Learning Model for Predicting Amputation Rate of Patients with Diabetic Foot Ulcer. Res. Sq. 2024. preprint. [Google Scholar] [CrossRef]
- Hong, S.; Chen, Y.; Lin, Y.; Xie, X.; Chen, G.; Xie, H.; Lu, W. Personalized prediction of diabetic foot ulcer recurrence in elderly individuals using machine learning paradigms. Technol. Health Care 2024, 32, 265–276. [Google Scholar] [CrossRef]
- Mousa, K.M.; Mousa, F.A.; Mohamed, H.S.; Elsawy, M.M. Prediction of foot ulcers using artificial intelligence for diabetic patients at Cairo university hospital, Egypt. SAGE Open Nurs. 2023, 9, 23779608231185873. [Google Scholar] [CrossRef]
- Zhang, D.; Dong, W.; Guan, H.; Yakupu, A.; Wang, H.; Chen, L.; Lu, S.; Tang, J. CT-Angiography-based outcome prediction on diabetic foot ulcer patients: A statistical learning approach. Diagnostics 2022, 12, 1076. [Google Scholar] [CrossRef] [PubMed]
- Mahbod, A.; Schaefer, G.; Ecker, R.; Ellinger, I. Automatic foot ulcer segmentation using an ensemble of convolutional neural networks. In Proceedings of the 2022 26th International Conference on Pattern Recognition (ICPR), Montreal, QC, Canada, 21–25 August 2022; pp. 4358–4364. [Google Scholar]
- Bouallal, D.; Douzi, H.; Harba, R. Diabetic foot thermal image segmentation using Double Encoder-ResUnet (DE-ResUnet). J. Med. Eng. Technol. 2022, 46, 378–392. [Google Scholar] [CrossRef]
- Thotad, P.N.; Bharamagoudar, G.R.; Anami, B.S. Diabetic foot ulcer detection using deep learning approaches. Sens. Int. 2023, 4, 100210. [Google Scholar] [CrossRef]
- Giridhar, C.; Akhila, B.; Kumar, S.P.; Sumalata, G. Detection of Multi Stage Diabetes Foot Ulcer using Deep Learning Techniques. In Proceedings of the 2024 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC), Salem, India, 5–7 June 2024; pp. 553–560. [Google Scholar]
- Jishnu, P.; BK, S.K.; Jayaraman, S. Automatic foot ulcer segmentation using conditional generative adversarial network (AFSegGAN): A wound management system. PLOS Digit. Health 2023, 2, e0000344. [Google Scholar]
- Basiri, R.; Manji, K.; Harton, F.; Poonja, A.; Popovic, M.R.; Khan, S.S. Synthesizing diabetic foot ulcer images with diffusion model. arXiv 2023, arXiv:2310.20140. [Google Scholar]
- Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. bmj 2021, 372, n71. [Google Scholar] [CrossRef]
- International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY). Available online: https://inplasy.com/ (accessed on 12 January 2025).
- Team, T. EndNote; Clarivate: Philadelphia, PA, USA, 2013; Volume 12. [Google Scholar]
- Goyal, M.; Reeves, N.D.; Davison, A.K.; Rajbhandari, S.; Spragg, J.; Yap, M.H. Dfunet: Convolutional neural networks for diabetic foot ulcer classification. IEEE Trans. Emerg. Top. Comput. Intell. 2018, 4, 728–739. [Google Scholar] [CrossRef]
- Amin, J.; Sharif, M.; Anjum, M.A.; Khan, H.U.; Malik, M.S.A.; Kadry, S. An integrated design for classification and localization of diabetic foot ulcer based on CNN and YOLOv2-DFU models. IEEE Access 2020, 8, 228586–228597. [Google Scholar] [CrossRef]
- Goyal, M.; Reeves, N.D.; Rajbhandari, S.; Ahmad, N.; Wang, C.; Yap, M.H. Recognition of ischaemia and infection in diabetic foot ulcers: Dataset and techniques. Comput. Biol. Med. 2020, 117, 103616. [Google Scholar] [CrossRef] [PubMed]
- Cassidy, B.; Reeves, N.D.; Pappachan, J.M.; Gillespie, D.; O’Shea, C.; Rajbhandari, S.; Maiya, A.G.; Frank, E.; Boulton, A.J.; Armstrong, D.G. The DFUC 2020 dataset: Analysis towards diabetic foot ulcer detection. Touchreviews Endocrinol. 2021, 17, 5. [Google Scholar] [CrossRef]
- Toofanee, M.S.A.; Dowlut, S.; Hamroun, M.; Tamine, K.; Petit, V.; Duong, A.K.; Sauveron, D. Dfu-siam a novel diabetic foot ulcer classification with deep learning. IEEE Access 2023, 11, 98315–98332. [Google Scholar] [CrossRef]
- Yap, M.H.; Cassidy, B.; Pappachan, J.M.; O’Shea, C.; Gillespie, D.; Reeves, N.D. Analysis towards classification of infection and ischaemia of diabetic foot ulcers. In Proceedings of the 2021 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI), Virtual Conference, 27–30 July 2021; pp. 1–4. [Google Scholar]
- Khalil, M.; Naeem, A.; Naqvi, R.A.; Zahra, K.; Muqarib, S.A.; Lee, S.-W. Deep learning-based classification of abrasion and ischemic diabetic foot sores using camera-captured images. Mathematics 2023, 11, 3793. [Google Scholar] [CrossRef]
- Bio. Diabetic Foot Ulcer (DFU). Available online: https://www.kaggle.com/datasets/laithjj/diabetic-foot-ulcer-dfu (accessed on 7 October 2024).
- Alqahtani, A.; Alsubai, S.; Rahamathulla, M.P.; Gumaei, A.; Sha, M.; Zhang, Y.-D.; Khan, M.A. Empowering foot health: Harnessing the adaptive weighted sub-gradient convolutional neural network for diabetic foot ulcer classification. Diagnostics 2023, 13, 2831. [Google Scholar] [CrossRef]
- Sathya Preiya, V.; Kumar, V.A. Deep Learning-Based Classification and Feature Extraction for Predicting Pathogenesis of Foot Ulcers in Patients with Diabetes. Diagnostics 2023, 13, 1983. [Google Scholar] [CrossRef]
- Fadhel, M.A.; Alzubaidi, L.; Gu, Y.; Santamaría, J.; Duan, Y. Real-time diabetic foot ulcer classification based on deep learning & parallel hardware computational tools. Multimed. Tools Appl. 2024, 83, 70369–70394. [Google Scholar]
- Gudivaka, R.K.; Gudivaka, R.L.; Gudivaka, B.R.; Basani, D.K.R.; Grandhi, S.H.; khan, F. Diabetic foot ulcer classification assessment employing an improved machine learning algorithm. Technol. Health Care 2025, 09287329241296417. [Google Scholar] [CrossRef] [PubMed]
- Lin, C.; Yuan, Y.; Ji, L.; Yang, X.; Yin, G.; Lin, S. The amputation and survival of patients with diabetic foot based on establishment of prediction model. Saudi J. Biol. Sci. 2020, 27, 853–858. [Google Scholar] [CrossRef]
- Schäfer, Z.; Mathisen, A.; Svendsen, K.; Engberg, S.; Rolighed Thomsen, T.; Kirketerp-Møller, K. Toward machine-learning-based decision support in diabetes care: A risk stratification study on diabetic foot ulcer and amputation. Front. Med. 2021, 7, 601602. [Google Scholar] [CrossRef]
- Wallach Kildemoes, H.; Toft Sørensen, H.; Hallas, J. The Danish national prescription registry. Scand. J. Public Health 2011, 39, 38–41. [Google Scholar] [CrossRef]
- Reddy, S.S.; Mahesh, G.; Preethi, N.M. Exploiting machine learning algorithms to diagnose foot ulcers in diabetic patients. EAI Endorsed Trans. Pervasive Health Technol. 2021, 7, e2. [Google Scholar] [CrossRef]
- Dong, W.; Nie, L.-J.; Wu, M.-J.; Xie, T.; Liu, Y.-K.; Tang, J.-J.; Dong, J.-Y.; Qing, C.; Lu, S.-L. WoundCareLog APP-a new application to record wound diagnosis and healing. Chin. J. Traumatol. 2019, 22, 296–299. [Google Scholar] [CrossRef]
- Popa, A.D.; Gavril, R.S.; Popa, I.V.; Mihalache, L.; Gherasim, A.; Niță, G.; Graur, M.; Arhire, L.I.; Niță, O. Survival prediction in diabetic foot ulcers: A machine learning approach. J. Clin. Med. 2023, 12, 5816. [Google Scholar] [CrossRef]
- Kaushal, R.K.; Pagidimalla, P.P.; Nalini, C.; Kumar, D. Predicting and Propagation of Diabetic Foot Infection by Deep Learning Model. EAI Endorsed Trans. Pervasive Health Technol. 2024, 10. [Google Scholar] [CrossRef]
- Tian, Z.; Wang, D.; Sun, X.; Cui, C.; Wang, H. Predicting the diabetic foot in the population of type 2 diabetes mellitus from tongue images and clinical information using multi-modal deep learning. Front. Physiol. 2024, 15, 1473659. [Google Scholar] [CrossRef]
- Arteaga-Marrero, N.; Hernandez, A.; Villa, E.; Gonzalez-Perez, S.; Luque, C.; Ruiz-Alzola, J. Segmentation approaches for diabetic foot disorders. Sensors 2021, 21, 934. [Google Scholar] [CrossRef] [PubMed]
- Huang, H.-N.; Zhang, T.; Yang, C.-T.; Sheen, Y.-J.; Chen, H.-M.; Chen, C.-J.; Tseng, M.-W. Image segmentation using transfer learning and Fast R-CNN for diabetic foot wound treatments. Front. Public Health 2022, 10, 969846. [Google Scholar] [CrossRef] [PubMed]
- Thomas, S. Medetec Wound Database: Stock Pictures of Wounds. Available online: http://www.medetec.co.uk/files/medetec-image-databases.html (accessed on 16 October 2024).
- Wang, C.; Anisuzzaman, D.; Williamson, V.; Dhar, M.K.; Rostami, B.; Niezgoda, J.; Gopalakrishnan, S.; Yu, Z. Fully automatic wound segmentation with deep convolutional neural networks. Sci. Rep. 2020, 10, 21897. [Google Scholar] [CrossRef]
- Wang, C.; Mahbod, A.; Ellinger, I.; Galdran, A.; Gopalakrishnan, S.; Niezgoda, J.; Yu, Z. FUSeg: The foot ulcer segmentation challenge. Information 2024, 15, 140. [Google Scholar] [CrossRef]
- Lan, T.; Li, Z.; Chen, J. FusionSegNet: Fusing global foot features and local wound features to diagnose diabetic foot. Comput. Biol. Med. 2023, 152, 106456. [Google Scholar] [CrossRef]
- Cassidy, B.; Kendrick, C.; Reeves, N.D.; Pappachan, J.M.; O’Shea, C.; Armstrong, D.G.; Yap, M.H. Diabetic foot ulcer grand challenge 2021: Evaluation and summary. In Diabetic Foot Ulcers Grand Challenge; Springer: Berlin/Heidelberg, Germany, 2021; pp. 90–105. [Google Scholar]
- Dhar, M.K.; Zhang, T.; Patel, Y.; Gopalakrishnan, S.; Yu, Z. FUSegNet: A deep convolutional neural network for foot ulcer segmentation. Biomed. Signal Process. Control 2024, 92, 106057. [Google Scholar] [CrossRef]
- Hresko, D.J.; Drotar, P.; Ngo, Q.C.; Kumar, D.K. Enhanced Domain Adaptation for Foot Ulcer Segmentation Through Mixing Self-Trained Weak Labels. J. Imaging Inform. Med. 2024, 1–12. [Google Scholar] [CrossRef]
- Kendrick, C.; Cassidy, B.; Reeves, N.D.; Pappachan, J.M.; O’Shea, C.; Chandrabalan, V.; Yap, M.H. Diabetic Foot Ulcer Grand Challenge 2022 Summary. In Diabetic Foot Ulcers Grand Challenge; Springer: Berlin/Heidelberg, Germany, 2022; pp. 115–123. [Google Scholar]
- Aliahmad, B.; Tint, A.N.; Poosapadi Arjunan, S.; Rani, P.; Kumar, D.K.; Miller, J.; Zajac, J.D.; Wang, G.; Ekinci, E.I. Is thermal imaging a useful predictor of the healing status of diabetes-related foot ulcers? A pilot study. J. Diabetes Sci. Technol. 2019, 13, 561–567. [Google Scholar] [CrossRef]
- Khandakar, A.; Chowdhury, M.E.; Reaz, M.B.I.; Ali, S.H.M.; Hasan, M.A.; Kiranyaz, S.; Rahman, T.; Alfkey, R.; Bakar, A.A.A.; Malik, R.A. A machine learning model for early detection of diabetic foot using thermogram images. Comput. Biol. Med. 2021, 137, 104838. [Google Scholar] [CrossRef]
- Hernandez-Contreras, D.A.; Peregrina-Barreto, H.; de Jesus Rangel-Magdaleno, J.; Renero-Carrillo, F.J. Plantar thermogram database for the study of diabetic foot complications. IEEE Access 2019, 7, 161296–161307. [Google Scholar] [CrossRef]
- Yogapriya, J.; Chandran, V.; Sumithra, M.; Elakkiya, B.; Shamila Ebenezer, A.; Suresh Gnana Dhas, C. Automated detection of infection in diabetic foot ulcer images using convolutional neural network. J. Healthc. Eng. 2022, 2022, 2349849. [Google Scholar] [CrossRef]
- Sarmun, R.; Chowdhury, M.E.; Murugappan, M.; Aqel, A.; Ezzuddin, M.; Rahman, S.M.; Khandakar, A.; Akter, S.; Alfkey, R.; Hasan, A. Diabetic Foot Ulcer Detection: Combining Deep Learning Models for Improved Localization. Cogn. Comput. 2024, 16, 1413–1431. [Google Scholar] [CrossRef]
- Solovyev, R.; Wang, W.; Gabruseva, T. Weighted boxes fusion: Ensembling boxes from different object detection models. Image Vis. Comput. 2021, 107, 104117. [Google Scholar] [CrossRef]
- Biswas, S.; Mostafiz, R.; Uddin, M.S.; Paul, B.K. XAI-FusionNet: Diabetic foot ulcer detection based on multi-scale feature fusion with explainable artificial intelligence. Heliyon 2024, 10, e31228. [Google Scholar] [CrossRef] [PubMed]
- Hyun, J.; Lee, Y.; Son, H.M.; Lee, S.H.; Pham, V.; Park, J.U.; Chung, T.-M. Synthetic data generation system for AI-based diabetic foot diagnosis. SN Comput. Sci. 2021, 2, 345. [Google Scholar] [CrossRef]
- Dua, D.; Graff, C. UCI machine learning repository, 2017, 7, 62. Available online: http://archive.ics.uci.edu/ml (accessed on 23 November 2024).
- Foomani, F.H.; Anisuzzaman, D.; Niezgoda, J.; Niezgoda, J.; Guns, W.; Gopalakrishnan, S.; Yu, Z. Synthesizing time-series wound prognosis factors from electronic medical records using generative adversarial networks. J. Biomed. Inform. 2022, 125, 103972. [Google Scholar] [CrossRef]
- Goyal, M.; Reeves, N.D.; Rajbhandari, S.; Yap, M.H. Robust methods for real-time diabetic foot ulcer detection and localization on mobile devices. IEEE J. Biomed. Health Inform. 2018, 23, 1730–1741. [Google Scholar] [CrossRef]
- Anandakrishan, K.; Jayaraman, S. Wound care: Wound segmentation and parameter estimation. In Proceedings of the 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI), Cartagena, Colombia, 18–21 April 2023; pp. 1–4. [Google Scholar]
- El-Kady, A.M.; Abbassy, M.M.; Ali, H.H.; Ali, M.F. Advancing Diabetic Foot Ulcer Detection Based On Resnet And Gan Integration. J. Theor. Appl. Inf. Technol. 2024, 102, 2258–2268. [Google Scholar]
- National Institute Of Diabetes And Endocrinology. Available online: https://nide.gov.eg/ (accessed on 6 May 2023).
- Chan, K.S.; Chan, Y.M.; Tan, A.H.M.; Liang, S.; Cho, Y.T.; Hong, Q.; Yong, E.; Chong, L.R.C.; Zhang, L.; Tan, G.W.L. Clinical validation of an artificial intelligence-enabled wound imaging mobile application in diabetic foot ulcers. Int. Wound J. 2022, 19, 114–124. [Google Scholar] [CrossRef]
- Anisuzzaman, D.; Patel, Y.; Niezgoda, J.A.; Gopalakrishnan, S.; Yu, Z. A mobile app for wound localization using deep learning. IEEE Access 2022, 10, 61398–61409. [Google Scholar] [CrossRef]
- AZH Wound Database. Available online: https://github.com/uwm-bigdata/wound_localization (accessed on 19 December 2024).
- Liu, W.; Anguelov, D.; Erhan, D.; Szegedy, C.; Reed, S.; Fu, C.-Y.; Berg, A.C. Ssd: Single shot multibox detector. In Proceedings of the Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, 11–14 October 2016; Proceedings, Part I 14. pp. 21–37. [Google Scholar]
- Cassidy, B.; Yap, M.H.; Pappachan, J.M.; Ahmad, N.; Haycocks, S.; O’Shea, C.; Fernandez, C.J.; Chacko, E.; Jacob, K.; Reeves, N.D. Artificial intelligence for automated detection of diabetic foot ulcers: A real-world proof-of-concept clinical evaluation. Diabetes Res. Clin. Pract. 2023, 205, 110951. [Google Scholar] [CrossRef] [PubMed]
- Sendilraj, V.; Pilcher, W.; Choi, D.; Bhasin, A.; Bhadada, A.; Bhadadaa, S.K.; Bhasin, M. DFUCare: Deep learning platform for diabetic foot ulcer detection, analysis, and monitoring. Front. Endocrinol. 2024, 15, 1386613. [Google Scholar] [CrossRef] [PubMed]
- Pappachan, J.M.; Cassidy, B.; Fernandez, C.J.; Chandrabalan, V.; Yap, M.H. The role of artificial intelligence technology in the care of diabetic foot ulcers: The past, the present, and the future. World J. Diabetes 2022, 13, 1131. [Google Scholar] [CrossRef] [PubMed]
- Liu, H.; Sun, W.; Cai, W.; Luo, K.; Lu, C.; Jin, A.; Zhang, J.; Liu, Y. Current status, challenges, and prospects of artificial intelligence applications in wound repair theranostics. Theranostics 2025, 15, 1662–1688. [Google Scholar] [CrossRef]
- Reifs-Jiménez, D.; Casanova-Lozano, L.; Reig-Bolañoo, R.; Grau-Carrión, S. Artificial Intelligence Methods to Assess the Decision Diagnosis in Chronic Wounds: A Systematic Review. J. Med. Syst. 2024. [Google Scholar] [CrossRef]
- Biswas, S.; Mostafiz, R.; Paul, B.K.; Uddin, K.M.M.; Hadi, M.A.; Khanom, F. DFU_XAI: A deep learning-based approach to diabetic foot ulcer detection using feature explainability. Biomed. Mater. Devices 2024, 2, 1225–1245. [Google Scholar] [CrossRef]
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[13] | 2020 | Classify DFUs and healthy skin | Image | DFU_QUTNet + SVM | Not provided | Batch size: 32, learning rate: 0.5, dropout rate: 0.5, number of layers: 58 | Training: 80%, testing: 20% | F1 score: 1. DFU_QUTNet: 94.5% 2. AlexNet: 89.1% 3. VGG16: 90.9% 4. GoogleNet: 92.9% 5. DFUNet: 93.1% |
[30] | 2020 | Classification and localization of ischemia and infection in DFUs | Image | CNN, YOLOv2 | Not provided | Learning rate: 0.001, epochs: 40 | Training: 50%, testing: 50% | Accuracy: Ischemia: 97% Infection: 99% |
[15] | 2023 | Classify ischemia and infection in DFUs | Image | ResNet50 | Not provided | Batch size: 32, learning rate: 0.001, momentum: 0.8, epochs: 30 | Training: 80%, validation: 10%, testing: 10% | Accuracy: 1. ResNet50: Ischemia: 99.49% Infection: 84.76% 2. AlexNet: Ischemia: 83.56% Infection: 83.22% 3. VGG16: Ischemia: 98.58% Infection: 79.32% 4. GoogleNet: Ischemia: 99.65% Infection: 79.66% 5. DenseNet: Ischemia: 99.30% Infection: 83.20% AUC: ResNet50: Ischemia: 99.96% Infection: 94.16% |
[33] | 2023 | Classify none, infection, ischemia, or both in DFUs | Image | DFU-SIAM + KNN | Not provided | Batch size: 8, learning rate: 10 × 10−6, epochs: 40 | K-fold validation (K = 5) | Accuracy: 95% |
[14] | 2023 | Classify ischemia and infection in DFUs | Image | EfficientNet | Not provided | Batch size, learning rate | Training: 70%, validation: 15%, testing: 15%, 5-fold cross-validation, transfer learning | Accuracy: 1. EfficientNet: Infection: 97% Ischemia: 99% 2. DenseNet121: Infection: 94% Ischemia: 96% 3. ResNet50: Infection: 87% Ischemia: 94% 4. Inception V3: Infection: 86% Ischemia: 98% 5. VGG16: Infection: 85% Ischemia: 95% |
[35] | 2023 | Classify abrasion foot sores and ischemic diabetic foot sores | Image | Vgg-19 with UNet++ | Not provided | Batch size: 32, learning rate: 0.05 (decayed by 0.1 every 10 epochs), momentum: 0.8, epochs: 150 | Training: 70%, validation: 20%, testing: 10%, transfer learning | 1. Vgg-19 with UNet++: Accuracy: 99.05% F1 score: 99.04% AUC: 0.996 2. Inception-v3: Accuracy: 95.52% F1 score: 95.53% AUC: 0.930 3. MobileNet: Accuracy: 96.73% F1 score: 96.94% AUC: 0.990 |
[37] | 2023 | Classify DFUs as normal or abnormal | Image | AWSg-CNN | Not provided | Batch size, learning rate | Training: 80%, testing: 20% | Accuracy, F1 score, AUC: 99% |
[38] | 2023 | Classify foot ulcer images as normal or abnormal | Image | DRNN, PFCNN | Not provided | Batch size: 32, learning rate: 0.01 (decayed by 0.1 every 25 epochs), momentum: 0.9, epochs: 2000 | Training: 50%, testing: 50% | Accuracy: 99.32% |
[39] | 2024 | Real-time classification of DFUs | Image | DFU_FNet, DFU_TFNet | Not provided | Batch size: 32, learning rate: 0.001, epochs: 100 | Transfer learning | 1. AlexNet: Accuracy: 89.11% F1 score: 88.1% 2. VGG16: Accuracy: 90.37% F1 score: 90.9% 3. GoogleNet: Accuracy: 91.93% F1 score: 92.9%. 4. DFU_FNet + SVM: Accuracy: 94.71% F1 score: 94.5% 5. DFU_TFNet: Accuracy: 99.81% F1 score: 99.25% |
[40] | 2025 | DFU classification | Image | RL, CPPN, SVM, ELM, ResNet50 | Not provided | Learning rate: 0.1, discount factor = 0.9 | Reinforcement learning | Classification accuracy: 93.75% Clustering Efficiency: Cluster 1: 71–88% Cluster 2: 85–97% Cluster 3: 90–98% Cluster 4: 93.5–98.2% |
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[41] | 2020 | Predict amputation and survival risks in patients with DF | Numerical | COX Regression-Based, BPNN, BPNN+GA | Not provided | Learning rate: 0.001, transfer function: trainlm and purelin | Training set: 160 patients; test set: 40 patients | AUC: 1. COX Regression: Amputation: 0.557 Death: 0.635 2. BPNN: Amputation: 0.924 Death: 0.712 3. BPNN+GA: Amputation: 0.891 Death: 0.712 |
[42] | 2021 | Predict risk of DFU and amputation | Numerical | LR, RF | Not provided | Not mentioned | Training: 75%, testing: 25% | Accuracy: LR: 95% |
[44] | 2021 | Predict presence of DFUs in patients | Numerical | ELM, SVM, ANN | Not provided | Hidden neurons: 35 (ELM), 10 (ANN); K value: 9 (KNN); SVM kernel: Gaussian | Training: 80%, testing: 20% | Accuracy: 1. ELM: 96.15% 2. SVM: 92.31% 3. ANN: 84.62% |
[19] | 2022 | Predict prognosis of DFUs | Numerical | ANN, LR | Not provided | ANN: Learning rate: 0.4, max training time: 15 min, hidden layer: hyperbolic tangent function, output layer: SoftMax function | Training: 60%, testing: 20%, holdout sample: 20% | 1. ANN: Accuracy: 91.6% AUC: 0.955 2. LR: Accuracy: 82.2% AUC: 0.874 |
[18] | 2023 | Predict positive or negative DFUs | Image and Numerical | ANN and DT | Not provided | ANN: Input features: 19, Hidden layers: 2 | Training: 75%, testing: 25%, k-fold cross-validation (k = 10) | Accuracy 1. ANN: 97% 2. DT: 93% |
[46] | 2023 | Predict mortality in patients with DFU | Numerical | MLP | Not provided | 5-year model: 3 hidden layers (4-1-5 neurons); 10-year model: 3 hidden layers (5-5-4 neurons) | Training: 80%, testing: 20%, 10-fold cross-validation | 1. 5-Year Model: Training Accuracy: 77.2% Test Accuracy: 72.4% 2. 10-Year Model: Training Accuracy: 75.9% Test Accuracy: 70.9% |
[47] | 2024 | Predict diabetic foot infections and ulcers | Numerical | Neural networks, DT, RF, regression model | Not provided | Not mentioned | Model comparison using multiple classifiers | 1. RF: MAE: 0.0077 RMSE: 0.0233 RAE: 1.6878% 2. DT: MAE: 0.0005 RMSE: 0.0040 RAE: 0.0800% 3. Neural Networks: MAE: 0.0065 RMSE: 0.0310 RAE: 1.4285% 4. Regression Model: MAE: 0.0144 RMSE: 0.023 RAE: 3.1635% |
[16] | 2024 | Predict amputation risk in patients with DFU | Numerical | SVM | Not provided | 5-fold cross-validation | Data are divided into training sets and test sets by using 5-fold cross-validation | Accuracy: 82.4% AUC: 0.89 |
[48] | 2024 | Predict DF in patients with T2DM using Traditional Chinese Medicine and Western medicine | Numerical and image | ResNet-50 | Not provided | Pretrained on ImageNet, last three convolution layers removed, FCL for feature extraction | 311 images for training, 80 for testing, 5 random seed sets | Accuracy: With tongue image 0.95 Without tongue image: 0.92 |
[17] | 2024 | Predict recurrence risk of DFUs in elderly diabetic patients | Numerical | SVM | Not provided | C = 20, degree = 2, gamma: scale, kernel type: linear | Training: 70%, testing: 30%, 5-fold cross-validation | Accuracy: SVM: 93% XGBoost: 82% KNN: 82% RF: 82% DT: 79% |
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[49] | 2021 | Evaluate segmentation methods for DF monitoring | Image | UPD, SPD, and SegNet | Not provided | Learning rate: 0.1, momentum: 0.9, epochs: 150 | Training: 50 images, testing: 24 images | 1. UPD: DICE: 95.35 ± 0.40% IoU: 91.11 ± 0.72% 2. SPD: DICE: 95.24 ± 0.52% IoU: 90.93 ± 0.95% 3. SegNet: DICE: 93.30 ± 2.91% IoU: 87.57 ± 5.01% |
[21] | 2022 | Segment diabetic foot images that fuse thermal and RGB data | Image | DE-ResUNet | Not provided | Learning rate: 0.01, momentum: 0.9, weight decay: 0.0005, ResNet-50 encoder | Training: 50%, validation: 25%, testing: 25% | IoU: 1. DE-ResUNet: 97% 2. SegNet: 95% 3. UNet: 95% |
[50] | 2022 | Image segmentation to enhance diagnosis of DFWs | Image | Fast R-CNN | Not provided | Not mentioned | Training: 30%, testing: 10%, transfer learning, training iterations: 100,000 | Accuracy: Wound image detection: 90% 1. Inception V2-coco: 87% 2. Kitti- ResNet101: 88% 3. Species-ResNet101: 89% |
[20] | 2022 | Segmentation of foot ulcers | Image | UNet and LinkNet | https://github.com/masih4/Foot_Ulcer_Segmentation (accessed on 18 November 2024) | Learning rate: 0.001 (reduced by 90% every 25 epochs), batch size: 4, epochs: 80, loss function: dice and focal loss | 5-fold cross-validation | Dice Score: UNet and LinkNet: 88.80% UNet with ASPP: 82.29% |
[54] | 2023 | Diagnose DFUs by integrating segmentation of wound areas with classification | Image | FusionSegNet | Not provided | Learning rate: 1 × 10−5, batch size: 16, epochs: 100, loss function: binary cross-entropy | 5-fold cross-validation | Accuracy: 1. FusionSegNet: 95.78% 2. Inception-ResNet-v2: 84.96% 3. DFUNet: 86.45% |
[56] | 2024 | Segmentation of DFUs | Image | FUSegNet | https://github.com/mrinal054/FUSegNet (accessed on 25 November 2024) | Learning rate: 1 × 10−4, batch size: 2, loss function: dice and focal loss, epochs: 200 | 5-fold cross-validation | Dice Score: 1. x-FUSegNet: 89% 2. UNet with HarDNet68: 87% 3. Stacked UNets: 86% |
[57] | 2024 | DFU segmentation using self-training and mixup augmentation | Image | Attention UNet | Not provided | Learning rate: 0.001, batch size: 8, epochs: 750, loss function: dice and cross-entropy | Training: 70%, validation: 10%, testing: 20%, 5-fold cross-validation | Dice Score: 1. DFUC 2022: 0.711 2. FUSeg: 0.859 |
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[60] | 2021 | Detection of DFUs | Image | AdaBoost, MobileNetV2, DenseNet201, ResNet50, InceptionV3 | Not provided | Transfer learning | Training: 80%, testing: 20%, 5-fold cross-validation | F1 Score: 1. AdaBoost: 97.75% 2. MobileNetV2: 92.50% 3. DenseNet201: 94.01% 4. ResNet50: 93.41% 5. InceptionV3: 93.71% |
[62] | 2022 | Detection of infection in DFU images | Image | DFINET | Not provided | Learning rate: 0.0001, batch size: 16, epochs: 30, loss function: binary cross-entropy | Training: 70%, validation: 20%, testing: 10% | F1 score: 1. DFINET: 92.12% 2. GoogLeNet: 76.39% 3. VGG16: 83.54% 4. AlexNet: 77.54% |
[22] | 2023 | Detection of DFUs | Image | EfficientNet | Not provided | EfficientNet (depth, width, resolution) | Training: 60%, validation: 20%, testing: 20% | F1 score: 1. AlexNet: 89.1% 2. VGG16: 91.0% 3. DFUNet: 93.3% 4. GoogleNet: 93.0% 5. EfficientNet: 99.0% |
[63] | 2024 | Detection and localization of DFUs | Image | YOLOv8m and Faster R-CNN ResNet101 | Not provided | Learning rate: 0.001, epochs: 100–150 | Training: 80%, validation: 10%, testing: 10%, transfer learning | F1 score: 0.780 mAP@0.5: 0.864 |
[65] | 2024 | Detection of DFUs using XAI | Image | FusionNet | Not provided | Learning rate: 0.0001, batch size: 32, epochs: 50, loss function: binary cross-entropy | Training: 70%, validation: 10%, eesting: 20%, transfer learning | 1. VGG19: Accuracy: 86.7% F1 Score: 85.4% 2. DenseNet201: Accuracy: 97.6% F1 Score: 97.7% 3. NASNetMobile: Accuracy: 77.3% F1 Score: 72.1% 4. FusionNet: Accuracy: 99.05% F1 Score: 99.08% |
[23] | 2024 | Detection and classification of DFUs | Image | DenseNet-201 | Not provided | Learning rate: 0.01, batch size: 30, epochs: 7 | Transfer learning | 1. DenseNet-201: Accuracy: 98% F1 score: None: 98% Infection: 97% Ischemia: 98% 2. EfficientNet-B3: Accuracy: 98% F1 score: None: 100% Infection: 97% Ischemia: 97% |
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[66] | 2021 | Synthetic data generation system | Numerical | NeuralProphet | Not provided | Epochs: 100 | Training: 60%, validation: 20%, testing: 20% | Accuracy of binary classification: 100% |
[68] | 2022 | Address limited EMR access by generating synthetic data | Numerical | EMR-TCWGAN | Not provided | Batch normalization | 4-fold cross-validation | AUC: 0.875 Accuracy: 77.98% |
[25] | 2023 | Generate synthetic DFU images | Image | Diffusion model | Not provided | Learning rate: 1 × 10−4 (decaying), batch size: 32, epochs: 500 | Gaussian noise | FID: 0.73 KID: 0.14 |
[24] | 2023 | Automatic foot ulcer segmentation | Image | AFSegGAN | Not provided | Learning rate: 0.002, batch size: 16 | Training steps = epochs × batch per epoch | IoU: 1. AFSegGAN: 99.07% 2. UNet-EffB2: 85.01% 3. DeepLabV3+SE: 92.4% |
[71] | 2024 | Enhance accuracy of DFU diagnosis | Image | ResNet50, ResNet50-GAN | Not provided | Learning rate: 0.001, batch size: 4 (ResNet50), 8 (ResNet50-GAN), epochs: 100 | 8-fold cross-validation | Accuracy: ResNet50: 76% ResNet50-GAN: 84% F1 score: ResNet50: 75% ResNet50-GAN: 84% |
Ref. | Year | Study Aim | Data Type | Model Used | Code Availability | Hyperparameter | Training Protocols | Evaluation Metrics |
---|---|---|---|---|---|---|---|---|
[6] | 2020 | Predict healing outcomes of DFUs | Numerical | RF, SVM | Not provided | RF: 2000 hyperparameter combinations (bootstrapping, tree criteria, depth, splits, PCA components); SVM: 2028 combinations (C, gamma, PCA components) | Training: 75%, testing: 25%, 3-fold cross-validation | Handcrafted image features: 1. RF: Accuracy: 0.811 F1 Score: 0.760 2. SVM: Accuracy: 0.784 F1 Score: 0.794 |
[73] | 2021 | Validate reliability of CARES4WOUNDS for wound measurement DFUs | Image and Numerical | C4W app | Not provided | Not mentioned | manual vs. AI-based wound measurement | Intra-rater reliability > 0.9 |
[74] | 2022 | Wound localization system in DFUs | Image | YOLOv3, tiny-YOLOv3 | https://github.com/uwm-bigdata/wound_localization (accessed on 19 December 2024) | Learning rate: 0.001, batch size: 8, epochs: 273 | Training: 4050 images | F1 Score: 0.949 |
[77] | 2023 | Automated detection of DFUs | Image | Faster R-CNN with Inception-ResNetV2 | Not provided | Transfer learning | Training: 1775 images | F1 Score: 94% |
[78] | 2024 | Detection, classification, and monitoring of DFUs | Image | YOLOv5s (localization), InceptionResNetV2 (classification) | Not provided | YOLOv5s: learning rate: 0.01, epochs: 30 | Training: 1800 images, testing: 200 images, 10-fold cross-validation | Classification accuracy: Infection: 79.76% Ischemic: 94.81% |
Metric | Equation | Parameters |
---|---|---|
Accuracy | Accuracy = | TP: True Positive, TN: True Negative, FP: False Positive, FN: False Negative |
F1 Score | F1 Score = 2 × | , |
Dice | Dice image = | TP: True Positive, FP: False Positive, FN: False Negative |
IoU | IoU image = | TP: True Positive, FP: False Positive, FN: False Negative |
AUC | AUC = | TP: True Positive, TN: True Negative, FP: False Positive, FN: False Negative |
MAE | MAE = | actual value, : predicted value, n: total number of data points |
RMSE | RMSE = 2 | actual value, : predicted value, n: total number of data points |
RAE | RAE = , | actual value, : predicted value, n: total number of data points, : mean of all actual values |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Alkhalefah, S.; AlTuraiki, I.; Altwaijry, N. Advancing Diabetic Foot Ulcer Care: AI and Generative AI Approaches for Classification, Prediction, Segmentation, and Detection. Healthcare 2025, 13, 648. https://doi.org/10.3390/healthcare13060648
Alkhalefah S, AlTuraiki I, Altwaijry N. Advancing Diabetic Foot Ulcer Care: AI and Generative AI Approaches for Classification, Prediction, Segmentation, and Detection. Healthcare. 2025; 13(6):648. https://doi.org/10.3390/healthcare13060648
Chicago/Turabian StyleAlkhalefah, Suhaylah, Isra AlTuraiki, and Najwa Altwaijry. 2025. "Advancing Diabetic Foot Ulcer Care: AI and Generative AI Approaches for Classification, Prediction, Segmentation, and Detection" Healthcare 13, no. 6: 648. https://doi.org/10.3390/healthcare13060648
APA StyleAlkhalefah, S., AlTuraiki, I., & Altwaijry, N. (2025). Advancing Diabetic Foot Ulcer Care: AI and Generative AI Approaches for Classification, Prediction, Segmentation, and Detection. Healthcare, 13(6), 648. https://doi.org/10.3390/healthcare13060648