Abstract:Abstract: Cardiopulmonary exercise testing (CPET) is a comprehensive testing method for evaluating cardiovascular function, exercise capacity, and overall health status. CPET involves multiple time series data of complex physiological parameters. Due to high correlations between the diverse physiological parameters, individual differences (such as age and physical condition), and a lack of unified interpretation standards, the clinical application and promotion of the existing CPET technologies are challenging. By extracting and analyzing features from cardiopulmonary time series data, a machine learning-based CPET system could facilitate disease diagnosis, exercise state recognition, and the assessmen of various physiological indicators. Furthermore, CPET technology based on machine learning can assist doctors in diagnosis and is more effective than traditional interpretation methods. This paper reviews the recent advances in the application of machine learning to CPET, encompassing disease identification, physiological parameter prediction, prognosis evaluation, and automated determination of ventilatory thresholds. It highlights the importance of model interpretability in clinical settings. In response to challenges such as data heterogeneity, limited model generalizability, and insufficient clinical applicability, the paper advocates for enhanced data sharing, the development of interpretable models, and the implementation of remote intelligent analysis, aiming to facilitate the transition of CPET from empirical interpretation toward a data-driven, intelligent evaluation paradigm.
