Machine learning in the prediction of diabetic peripheral neuropathy: a systematic review

机器学习在糖尿病周围神经病变预测中的应用:系统性综述

阅读:2

Abstract

OBJECTIVE: This systematic review provides an overview of machine learning (ML) methods for predicting diabetic peripheral neuropathy (DPN). METHOD: We searched PubMed, Embase, Cochrane Library, and Web of Science databases with the search period limited from their inception to December 3, 2024 (the last search date). The search terms were restricted to “diabetes,” “neuropathy,” and “machine learning.” All studies that developed or validated prognostic models for DPN using ML were considered. Prediction model Risk of Bias ASsessment Tool (PROBAST) was used to assess the risk of bias and applicability of included studies. RESULTS: A total of 888 studies were retrieved and 15 articles were included. Most were retrospective studies, with sample sizes ranging from 90 to 102,876 patients. All 15 studies utilised internal validation methods, three studies employed both internal and external validation methods. Internal validation methods like cross-validation were widely used, with area under the curve (AUC) ranging from 0.640 to 0.900. A total of six studies reported complete AUC values yielding a pooled AUC of 0.773 (CI: 0.707–0.839, I²= 99.14). A total of 34 different ML algorithms were utilised across the studies, with the top five being logistic regression, random forest, support vector machine, decision tree, and XGBoost. Calibration was reported in 6 studies, showing satisfactory performance. All studies had a high risk of bias, but most models demonstrated good applicability. CONCLUSION: Existing DPN prediction models demonstrate good performance in discrimination. However, the evaluation indicates that the overall risk of bias in the included studies is high, and their applicability is limited. Future efforts should prioritize prospective, large, multicentre datasets, external validation, and adherence to PROBAST guidelines to reduce bias and enhance applicability for clinical application. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12911-025-03201-6.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。