Abstract
119 COPD patients (with no other pulmonary pathology, heart failure, renal failure or malignancy) and 63 healthy controls were recruited. Subjects underwent screening questions then performed spirometry, had BMI calculated, Oxygen saturations recorded and smoking status validated with eCO measurement. Subjects fasted for four hours and rested for 20 minutes in the same room prior to breath sample collection in triplicate.
Breath samples were collected using Bio-VOC® breath sampler and analyzed by Gas Chromatography-Mass Spectrometry. Multiple Machine learning approaches were applied on total VOCs (n= 2075) as well as pre-selected VOCs (n= 128, which only occurred in more than 5% of subjects with quality rating above 50%) to generate predictive models. Adequacy of models was assessed by generating Receiver Operating Characteristic (ROC) curves.
Breath samples were collected using Bio-VOC® breath sampler and analyzed by Gas Chromatography-Mass Spectrometry. Multiple Machine learning approaches were applied on total VOCs (n= 2075) as well as pre-selected VOCs (n= 128, which only occurred in more than 5% of subjects with quality rating above 50%) to generate predictive models. Adequacy of models was assessed by generating Receiver Operating Characteristic (ROC) curves.
Original language | English |
---|---|
Pages (from-to) | A4517-A4517 |
Number of pages | 1 |
Journal | American Journal of Respiratory and Critical Care Medicine |
Volume | 185 |
DOIs | |
Publication status | Published - May 2012 |