Automating the interpretation of cardiopulmonary exercise testing: a proof-of-concept study

Cardiopulmonary exercise testing (CPET) is a valuable tool to evaluate the response to exercise of the pulmonary, cardiovascular and muscular system. However, its use in clinical practice is hampered by the complexity of its interpretation. We aimed to verify if the development of a rule-based algorithm for automated interpretation of CPET is feasible. Using Python 3.12, we developed a rule-based algorithm that automatically interprets 9 aspects of a CPET. The implemented rules and thresholds were based on the results of a systematic search of existing literature and clinical expertise. As proof-of-concept, the algorithm was run on 30 randomly selected CPETs of patients with chronic respiratory disease (COPD, IPF and pulmonary hypertension). The outcomes of the algorithm were reviewed by 2 CPET experts: 1 was involved in the development of the algorithm (expert 1) and 1 was completely independent (expert 2). The degree of agreement (yes/no) with the algorithm’s outcome was recorded. Overall, agreement between the outcome of the algorithm and CPET experts was good to excellent, ranging from 87% to 100% across different CPET aspects for expert 1 and from 73% to 97% for expert 2 (table 1). Based on this preliminary investigation with a limited sample size, the automated interpretation of CPET by a rule-based algorithm seems feasible. Additional validation on a larger, more diverse sample by more CPET experts is necessary. <fig><object-id>erj;64/suppl_68/OA1962/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig>