Research advance and challenges of the application of machine learning in laboratory medicine / 中华检验医学杂志

ABSTRACT

The application of machine learning has become an important direction for the development of intelligent laboratory medicine. Recently, the rapid development of open-source software and publicly available data sources made the application of machine learning highly accessible. It reduced the requirement for developers to have necessary matter knowledge and also facilitated a surge in interest and publications. However, the practicality and reproducibility of machine learning models still remain unclear. In the face of these challenges, some countermeasures were proposed, including strict control of data quality, improvemrnt of model applicability, establishment of model selction and validation strategies, enhancement of model interpretability and reproducibility. Machine learning helps to break through the bottleneck of clinical translation of laboratory big data and improve the quality of diagnostic services in the laboratory medicine.