Systematic review of diabetic foot ulcer classification models using artificial intelligence and machine learning techniques (Preprint)

• Type: Article in International Scientific Journal
• Date: 2024

Authors

Silva, MA (Author) Other; Hamilton, EJ (Author) Other; Russell, DA (Author) Other; Game, F (Author) Other; Wang, SC (Author) Other; Baptista, S (Author) FMUP; Monteiro-Soares, M (Author) Other

Journal

Title: Journal of Medical Internet Research

ISSN: 1438-8871

Publisher: JMIR Publications

Other information

Authenticus ID: P-017-M1Z

DOI: 10.2196/69408

Abstract (EN): <sec> <title>BACKGROUND</title> <p>Diabetes-related foot ulceration (DFU) is a common complication of diabetes, with a significant impact on survival, healthcare costs and health-related quality of life.</p> </sec> <sec> <title>OBJECTIVE</title> <p>We aimed to identify and collect the available evidence assessing the ability of machine learning (ML) based models in predicting clinical outcomes in people with DFU.</p> </sec> <sec> <title>METHODS</title> <p>We searched the MEDLINE database (PubMed), Scopus, Web of Science and IEEExplore for articles published up to July 2023. Studies were eligible if they were anterograde analytical studies that examined the prognostic abilities of ML models in predicting clinical outcomes in a population that included at least 80% of adults with DFU. The literature was screened independently by two investigators for eligibility criteria and extracted data. The risk of bias was evaluated using the Quality In Prognosis Studies (QUIPS) tool and Prediction Model Risk Of Bias Assessment Tool (PROBAST) by two investigators independently.</p> </sec> <sec> <title>RESULTS</title> <p>We retrieved a total of 2412 references after removing the duplicates, of which 167 were subjected to full text screening. Two references were added from searching relevant studies¿ list of references. A total of 11 studies, comprising 13 articles, were included focusing on three outcomes: wound healing, lower extremity amputation and mortality. Overall, 55 predictive models were created using mostly clinical characteristics, random forest as the developing method and area under the curve (AUC) as discrimination accuracy measure. AUC varied from 0.56 to 0.94, with the majority of the models reporting an AUC ¿ 0.8. All studies were found to have a high risk of bias, mainly due to a lack of uniform variable definitions, outcome definitions and follow-up periods, insufficient sample sizes, and inadequate handling of missing data.</p> </sec> <sec> <title>CONCLUSIONS</title> <p>The ML-based models found in this study achieved good discrimination ability in people with DFU. However, models presented a high risk of bias meaning further studies with a stricter methodology are needed.</p> </sec> <sec> <title>CLINICALTRIAL</title> <p>PROSPERO CRD42022308248; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=308248</p> </sec>

Language: English

Type (Professor's evaluation): Scientific