Machine Learning Model Discriminate Ischemic Heart Disease Using Breathome Analysis

Abstract

Background: Ischemic heart disease (IHD) impacts the quality of life and is the most frequently reported cause of morbidity and mortality globally. Aims: To assess the changes in the exhaled volatile organic compounds (VOCs) in patients with vs. without ischemic heart disease (IHD) confirmed by stress computed tomography myocardial perfusion (CTP) imaging. Objectives: IHD early diagnosis and management remain underestimated due to the poor diagnostic and therapeutic strategies including the primary prevention methods. Materials and Methods: A single center observational study included 80 participants. The participants were aged ≥ 40 years and given an informed written consent to participate in the study and publish any associated figures. Both groups, G1 (n = 31) with and G2 (n = 49) without post stress-induced myocardial perfusion defect, passed cardiologist consultation, anthropometric measurements, blood pressure and pulse rate measurements, echocardiography, real time breathing at rest into PTR-TOF-MS-1000, cardio-ankle vascular index, bicycle ergometry, and immediately after performing bicycle ergometry repeating the breathing analysis into the PTR-TOF-MS-1000, and after three minutes from the end of the second breath, repeat the breath into the PTR-TOF-MS-1000, then performing CTP. LASSO regression with nested cross-validation was used to find the association between the exhaled VOCs and existence of myocardial perfusion defect. Statistical processing performed with R programming language v4.2 and Python v.3.10 [^R], STATISTICA program v.12, and IBM SPSS v.28. Results: The VOCs specificity 77.6% [95% confidence interval (CI); 0.666; 0.889], sensitivity 83.9% [95% CI; 0.692; 0.964], and diagnostic accuracy; area under the curve (AUC) 83.8% [95% CI; 0.73655857; 0.91493173]. Whereas the AUC of the bicycle ergometry 50.7% [95% CI; 0.388; 0.625], specificity 53.1% [95% CI; 0.392; 0.673], and sensitivity 48.4% [95% CI; 0.306; 0.657]. Conclusions: The VOCs analysis appear to discriminate individuals with vs. without IHD using machine learning models. Other: The exhaled breath analysis reflects the myocardiocytes metabolomic signature and related intercellular homeostasis changes and regulation perturbances. Exhaled breath analysis poses a promise result to improve the diagnostic accuracy of the physical stress tests using machine learning models.

Keywords

• breathome
• metabolome
• PTR-TOF-MS
• VOCs
• IHD
• optimizing