Advanced search
Publication Cover
Diabetes, Metabolic Syndrome and Obesity Volume 15, 2022 - Issue
Submit an article Journal homepage
415
Views
0
CrossRef citations to date
0
Altmetric
ORIGINAL RESEARCH

Machine Learning Models for Predicting the Risk of Hard-to-Heal Diabetic Foot Ulcers in a Chinese Population

Shiqi Wang1 Department of Burns and Plastic Surgery, Affiliated Drum Tower Hospital, Medical School of Nanjing University, Nanjing, People’s Republic of ChinaORCID Iconhttps://orcid.org/0000-0003-1991-8893View further author information
ORCID Icon,
Chao Xia2 Department of Orthopedics, Air Force Hospital of Eastern Theater Command, Nanjing, People’s Republic of ChinaView further author information
,
Qirui Zheng3 Software Institute, Nanjing University, Nanjing, People's Republic of ChinaView further author information
,
Aiping Wang4 Department of Endocrinology, Air Force Hospital of Eastern Theater Command, Nanjing, People's Republic of ChinaCorrespondence[email protected]
View further author information
&
Qian Tan1 Department of Burns and Plastic Surgery, Affiliated Drum Tower Hospital, Medical School of Nanjing University, Nanjing, People’s Republic of ChinaCorrespondence[email protected]
View further author information
Pages 3347-3359 | Received 01 Aug 2022, Accepted 20 Oct 2022, Published online: 29 Oct 2022

Figures & data

Figure 1 (A) Workflow of the study; (B) Flowchart of patient selection.

Abbreviations: AUC, area under the curve; AdaBoost, adaptive boosting; DFU, diabetic foot ulcer; GLM, generalized linear model; KNN, k-nearest neighbor; NB, naïve Bayes; PAR, percent area reduction; RF, random forest; SVM, support vector machine.
Figure 1 (A) Workflow of the study; (B) Flowchart of patient selection.

Table 1 Differences Between Demographic and Clinical Characteristics of Healed and Hard-to-Heal Groups

Table 2 Prediction Factors for Hard-to-Heal in DFUs

Figure 2 Confusion matrix of the risk prediction models with machine learning algorithms. (A) AdaBoost: adaptive boosting; (B) GLM: general linear regression; (C) KNN: k-nearest neighbor; (D) NB: naïve Bayes; (E) RF: random forest; (F) SVM: support vector machine.

Figure 2 Confusion matrix of the risk prediction models with machine learning algorithms. (A) AdaBoost: adaptive boosting; (B) GLM: general linear regression; (C) KNN: k-nearest neighbor; (D) NB: naïve Bayes; (E) RF: random forest; (F) SVM: support vector machine.

Table 3 The Comparisons of Machine Learning Algorithms

Figure 3 ROC curves for predicting hard-to-heal in DFU patients with machine learning algorithms.

Abbreviations: AdaBoost, adaptive boosting; AUC, area under the curve; DFUs, diabetic foot ulcers; GLM, general linear regression; KNN, k-nearest neighbor; NB, naïve Bayesian; RF, random forest; ROC, Receiver operating characteristic; SVM, support vector machine.
Figure 3 ROC curves for predicting hard-to-heal in DFU patients with machine learning algorithms.

Figure 4 The values of evaluation metrics of six machine learning algorithms.

Abbreviations: AdaBoost, adaptive boosting; AUC, area under the curve; GLM, general linear regression; KNN, k-nearest neighbor; NB, naïve Bayesian; RF, random forest; SVM, support vector machine.
Figure 4 The values of evaluation metrics of six machine learning algorithms.

Figure 5 Feature importance ranking of the included feature of the naïve Bayesian model.

Abbreviations: CRP, C-reactive protein.
Figure 5 Feature importance ranking of the included feature of the naïve Bayesian model.

Related research

People also read lists articles that other readers of this article have read.

Recommended articles lists articles that we recommend and is powered by our AI driven recommendation engine.

Cited by lists all citing articles based on Crossref citations.
Articles with the Crossref icon will open in a new tab.