Electronic Nose-Based Assessment of Cardiovascular Disease Severity Using Blood Analysis

An electronic nose-based approach for cardiovascular disease severity assessment is presented, leveraging blood samples instead of conventional breath analysis to improve standardization within clinical workflows. The proposed method extends current e-nose applications by enabling 3-class discrimination among control subjects and two levels of disease severity, allowing finer clinical stratification. Sensor responses are enhanced through the integration of optimized custom sensors and analyzed using an ensemble machine learning model. A realistic evaluation of performance is ensured by adopting a subject-independent validation strategy based on patient-level group-wise cross-validation, which prevents information leakage from repeated measurements of the same subject. Under these conditions, the system achieved an accuracy of 76% (± 15%). These results highlight the potential of combining advanced sensing technologies with rigorous validation protocols to develop reliable and clinically relevant e-nose systems for biomedical applications.