Diabetic foot ulcers (DFUs) are a major source of preventable morbidity. Best estimates place the lifetime risk of a foot ulcer at ~19–34% among people with diabetes. (1, 2)

These wounds often precede infection, amputation, and premature mortality. Contemporary reviews find 5-year mortality after major lower-extremity amputation is very high (about 40–70%), varying by population, comorbidity, and amputation level. (3, 4)

Early detection: more than a clinical convenience

Daily home skin-temperature monitoring is one of the best-studied early-warning strategies.

• In a randomized trial, adding home temperature checks to standard care reduced foot complications from 20% to 2% over ~6 months (complications included ulcers and Charcot neuroarthropathy). (5)
• A multicenter randomized trial in high-risk patients showed lower ulcer recurrence with temperature monitoring (8.5%) compared with standard therapy (29.3%) or structured foot inspection (30.4%) over 15 months. (6)
• A newer multicenter RCT found no significant reduction in ulcers on an intention-to-treat basis; however, among participants who reduced ambulatory activity when “hotspots” appeared, ulcer recurrence was significantly lower, underscoring that behavioral response to alerts is pivotal. (7)

AI in wound assessment: precision through data 

On standardized DFU image datasets/challenges:

• The DFU Segmentation Challenge reported a best mean Dice ≈ 0.729 for delineating ulcer regions—supporting automated size/area measurements. (8)
• The DFUC2020 detection challenge reported best mean average precision (mAP) ≈ 0.694 and F1 ≈ 0.743 for localizing ulcers in clinical photos. These are engineering benchmarks (not direct clinical outcomes) but show maturing reliability for consistent, objective measurements. (9)

Explainable AI and the impact of structured care

Recent DFU-focused work applies Grad-CAM and related XAI (explainable AI) tools to highlight the regions that drive a model’s prediction, improving interpretability for clinicians reviewing automated outputs (10, 11).

At the same time, access to specialized foot care appears to play a major role in real-world outcomes. In a U.S. Medicaid claims analysis (2010–2015), states covering podiatry services had 48% lower odds of major amputation (OR 0.52) and 24% lower odds of hospitalization for foot infection (OR 0.76) within 12 months of a new DFU diagnosis compared with states without such coverage. Minor amputations were more common (OR 1.58). As an observational study, this reflects association rather than causation, but it underscores the value of structured access and multidisciplinary care (12).

Achieving early detection with AI: A practical approach

For high-risk patients, the most evidence-consistent approach today is:

1. Daily self-monitoring (e.g., skin-temperature checks).
2. Actionable response (reducing activity/offloading when a hotspot is detected and contacting the care team).
3. Structured access to podiatric services and multidisciplinary care.
   AI-assisted image analysis can standardize measurements (size, boundaries, tissue changes) and augment clinical review, while explainability tools help clinicians validate algorithm focus before acting. (7, 8, 9, 11)

To learn more about our predictive health monitoring and AI-supported wound care for SNFs and assisted living, visit our services.

References

1. McDermott, K., et al. (2022). Etiology, epidemiology, and disparities in the burden of diabetic foot disease. Diabetes Care, 45(12), 2736–2749. https://pmc.ncbi.nlm.nih.gov/articles/PMC9797649/ 
2. Everett, E., & Mathioudakis, N. (2018). Update on management of diabetic foot ulcers. Annals of the New York Academy of Sciences, 1411(1), 153–165. https://pmc.ncbi.nlm.nih.gov/articles/PMC5793889/
3. Kurichi, J. E., & Kwong, P. L. (2023). Analysis of 5-year mortality following lower extremity amputation. Cureus, 15(1), e33373. https://pmc.ncbi.nlm.nih.gov/articles/PMC9833438/ 
4. Fortington, L. V., et al. (2013). Mortality after non-traumatic major amputation among patients with peripheral arterial disease and/or diabetes: A systematic review. Journal of Foot & Ankle Surgery. https://www.jfas.org/article/S1067-2516%2816%2900013-2/abstract
5. Lavery, L. A., et al. (2004). Home monitoring of foot skin temperatures to prevent ulceration. Diabetes Care, 27(11), 2642–2647. https://pubmed.ncbi.nlm.nih.gov/15504999/
6. Lavery, L. A., et al. (2007). Preventing diabetic foot ulcer recurrence in high-risk patients using temperature monitoring. Diabetes Care, 30(1), 14–20. https://pubmed.ncbi.nlm.nih.gov/17192326
7. Bus, S. A., et al. (2021). Effectiveness of at-home skin temperature monitoring in preventing diabetic foot ulcer recurrence: Randomized controlled trial. BMJ Open Diabetes Research & Care, 9(1), e002392. https://drc.bmj.com/content/9/1/e002392 
8. Yap, M. H., et al. (2024). Diabetic foot ulcers segmentation challenge: Benchmark and analysis. Medical Image Analysis. https://www.sciencedirect.com/science/article/pii/S1361841524000781 
9. Yap, M. H., et al. (2020). Deep learning in DFU detection: A comprehensive evaluation (DFUC2020). arXiv preprint. https://arxiv.org/abs/2010.03341 
10. Rathore, P. S., et al. (2025). Feature-explainability-based deep learning for DFU identification with Grad-CAM maps. Scientific Reports. https://www.nature.com/articles/s41598-025-90780-z 
11. Guan, H., et al. (2024). The role of machine learning in advancing diabetic foot. Frontiers in Endocrinology. https://pmc.ncbi.nlm.nih.gov/articles/PMC11089132/ 
12. Luu, I. Y., et al. (2024).mproved Diabetic Foot Ulcer Outcomes in Medicaid Beneficiaries with Podiatric Care Access. ClinicoEconomics and Outcomes Research. https://pmc.ncbi.nlm.nih.gov/articles/PMC11706342/