Quality assessment of machine learning models for diagnostic imaging in orthopaedics: A systematic review
作者：
Highlights：
• ML items researchers commonly fail to report are identified and recommendations for improvement are provided in this review.
• Few studies reported metrics of model performance such as the handling of missing data and data-preprocessing steps.
• High risk of bias regarding patient selection was found in 18% of studies and was undeterminable in 32% of included studies.
• We encourage authors to follow methodological guidelines to perpetuate transparent reporting
• Future investigations developing ML-based imaging models are encouraged to include explainability methods to identify bias
摘要
•ML items researchers commonly fail to report are identified and recommendations for improvement are provided in this review.•Few studies reported metrics of model performance such as the handling of missing data and data-preprocessing steps.•High risk of bias regarding patient selection was found in 18% of studies and was undeterminable in 32% of included studies.•We encourage authors to follow methodological guidelines to perpetuate transparent reporting•Future investigations developing ML-based imaging models are encouraged to include explainability methods to identify bias
论文关键词：Artificial intelligence,Machine learning,Medical imaging,Orthopaedics
论文评审过程：Received 29 March 2022, Revised 30 August 2022, Accepted 30 August 2022, Available online 6 September 2022, Version of Record 9 September 2022.
论文官网地址：https://doi.org/10.1016/j.artmed.2022.102396
