Predicting and Diagnosing Pneumonia in Patients Undergoing Elective Cardiac Surgery via Machine Learning Analysis of Exhaled Volatile Carbonyl Compounds

Toyokazu Endo; Kevin Tran; Dylan A Goodin; Gianna Katsaros; Zhenzhen Xie; Xiao-An Fu; George Pantalos; Hermann B Frieboes; Victor van Berkel

doi:10.1016/j.jtcvs.2025.06.028

Predicting and Diagnosing Pneumonia in Patients Undergoing Elective Cardiac Surgery via Machine Learning Analysis of Exhaled Volatile Carbonyl Compounds

J Thorac Cardiovasc Surg. 2025 Jul 2:S0022-5223(25)00548-3. doi: 10.1016/j.jtcvs.2025.06.028. Online ahead of print.

Authors

Toyokazu Endo¹, Kevin Tran², Dylan A Goodin², Gianna Katsaros¹, Zhenzhen Xie³, Xiao-An Fu³, George Pantalos¹, Hermann B Frieboes², Victor van Berkel⁴

Affiliations

¹ Department of Cardiovascular and Thoracic Surgery, University of Louisville School of Medicine.
² Department of Bioengineering, University of Louisville Speed School of Engineering.
³ Department of Chemical Engineering, University of Louisville Speed School of Engineering.
⁴ Department of Cardiovascular and Thoracic Surgery, University of Louisville School of Medicine. Electronic address: Victor.vanberkel@louisville.edu.

PMID: 40615109
DOI: 10.1016/j.jtcvs.2025.06.028

Abstract

Objective: Pneumonia remains one of the most common post-operative complications after elective cardiac surgery. Early intervention could lead to improved patient outcomes, including lower rates of ICU admissions, and shorter hospital stays. Volatile organic compounds (VOCs) in exhaled breath have shown promise in diagnosis and classification for various lung-related conditions. The study aims to diagnose and predict the onset of pneumonia in patients undergoing elective cardiac surgery via machine learning (ML) analysis of VOCs.

Methods: Patients undergoing elective cardiac surgery (n=75) were enrolled in the study (March 2023 - July 2024). Each patient's breath was collected in a 600mL Tedlar bag pre-operatively, within 24 hours, and every three days. The pneumonia group consisted of those who developed clinical signs of pneumonia post-operatively. Carbonyl compounds in the breath were captured on a microchip and identified using mass spectrometry. A machine learning (ML) workflow was implemented to build a model for pneumonia diagnosis (trained on pre- and post-operative VOC samples) and to build a prediction model of pneumonia development (trained on pre-operative samples) (alpha 0.05).

Results: Of the 75 patients enrolled during the study period, 10 developed clinical signs of pneumonia. The majority of patients had undergone coronary arterial bypass grafting (CABG) (50.1%), followed by aortic valve/root replacement (22.7%), concomitant CABG and valve (16%), and mitral valve repair/replacement (8%). Twenty-four carbonyls were selected by the pneumonia diagnosis model, including formaldehyde, hexanal, C₁₀H₂₀O, C₁₁H₂₂O, hexanone, and hydroxy-butanal. The proposed pneumonia diagnosis model had an area under the receiver oOperating cCharacteristic (AUROC) of 0.833 and an area under the precision-recall curve (PRAUC) of 0.818 on the test set. In contrast, four carbonyls (heptanal, octenone, C₁₂H₂₄O and acetone) were selected by the model to predict the onset of pneumonia using pre-operative breath samples (AUROC of 0.833 and PRAUC of 0.818 on the test set).

Conclusions: This pilot study demonstrates that VOCs captured from breath can be used to train and test ML models for diagnosis and prediction of pneumonia onset in patients undergoing elective cardiac surgery. This finding has implications for guiding perioperative and post-operative strategies for preventing pneumonia.

Keywords: Cardiac Surgery; Machine Learning; Pneumonia; Volatile Carbonyl Compounds.