Explainable machine-learning decision support for diabetic foot ulcers, externally validated and tested in a randomised comparison
An ensemble of gradient-boosted trees and neural networks trained on records of 763 diabetic foot ulcer patients at two Chinese hospitals predicted cure, recurrence, amputation and death at a third hospital with a macro-averaged AUC of 0.901; in a 202-patient randomised comparison of AI-assisted against usual care, the AI-assisted arm had better short-term follow-up engagement and glucose-management adherence, and blinded experts scored AI-assisted versions of usual-care recommendations higher than the clinicians' originals (7.77 against 6.47 out of 9).
- Model
- Ensemble of single-task XGBoost and multilayer perceptron models with SHAP explanations
- Field
- Medicine
- Date
- 2026-10-31
- Human collaborators
- Jia-Yu Zhang, Yangyang Zhuang, Qingping Cao, Aijun Zhang, Chi Zhang
What was found
Development used electronic records and follow-up from 2018 to 2025 at two hospitals in Jiangsu and Anhui, with a locked model then tested on 100 patients from a hospital in Shandong; discrimination was highest for amputation (AUC 0.953) and death (0.951) and lowest for recurrence (0.820). From January to March 2026, 202 patients were randomised 1:1. In the AI-assisted arm physicians drafted a plan without the tool, then reviewed the four predicted risks, SHAP explanations and AI-generated recommendations through a web platform before revising, and kept final authority; the tool added about four minutes of data entry and two minutes of review per encounter. In the usual-care arm physicians worked without it.
Novelty check
Europe PMC, 2026-11-12: "diabetic foot" AND ("artificial intelligence" OR "machine learning") AND (randomised OR randomized), 1,141 records, top 25 by relevance read: this paper, reviews and editorials, a 2025 protocol for a randomised trial of AI remote wound monitoring, and externally validated models predicting ulcer onset. No earlier completed randomised evaluation of AI decision support for diabetic foot ulcer prognosis was found among them, and no earlier entry in this registry covers diabetic foot care.
Caveats and known objections
The headline quality score is not a comparison between randomised arms. It comes from a nested paired evaluation: for the 100 usual-care cases, an AI-assisted version of each finalised clinician recommendation was generated offline, and the same expert scored both versions blind to source, one expert per document with no adjudication. The between-arm outcomes are follow-up engagement and adherence at 4 and 12 weeks, which are surrogates; 61 of 200 analysed patients were lost to follow-up and 86 responded at 12 weeks, and no difference in healing, amputation or death is reported, which the authors acknowledge. No trial registration number appears in the paper. Three hospitals in one country, 863 retrospective patients and an external validation set of 100. Funded by the Xuzhou Municipal Health Commission; the developers ran the trial, no commercial vendor is involved, the authors declare no competing interests, and the code is promised for public release. Autonomy is ai-assisted: physicians drafted plans first and made the final decision.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Explainable machine-learning decision support for diabetic foot ulcers, externally validated and tested in a randomised comparison. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-31-dfu-explainable-decision-support
BibTeX
@misc{whataifound-affiliatedhospitalofxuzhoumedicaluniversity-2026-support,
title = {Explainable machine-learning decision support for diabetic foot ulcers, externally validated and tested in a randomised comparison},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Affiliated Hospital of Xuzhou Medical University / Westlake University. Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-10-31-dfu-explainable-decision-support}
}