Identification of Lung Cancer Breath Biomarkers using Infrared Cavity Ring-Down Spectroscopy

Identification of Lung Cancer Breath Biomarkers using Infrared Cavity Ring-Down Spectroscopy

Robyn Larracy, University of New Brunswick 

Poster PDF

Abstract:

Lung cancer remains the most prevalent, deadliest form of cancer worldwide, underlining the need for regular screening. While low-dose computed tomography is currently used for this purpose, breath analysis has shown potential as a more effective, affordable, and accessible alternative. 
 
Our work considers a cavity ring-down spectroscopy (CRDS)-based system for differentiating lung cancer and non-lung cancer individuals through exhaled breath. Ten-litre alveolar breath samples were collected from 96 control subjects (45 males, 56 former or current smokers) and 62 pre-treatment, non-small cell lung cancer patients (31 males, 59 former or current smokers). CRDS analysis was performed at four different desorption temperatures (75°C, 150°C, 225°C, and 300°C) for each sample, yielding four absorption spectra for 73 CO2 laser lines in the mid-infrared range. A stepwise spectral fitting program based on a library of 152 volatile organic compounds (VOCs) was used to quantify the compounds presented in each spectrum.  
 
The VOC concentrations from each spectrum were used to build a classification system. The concentrations were log-transformed and meaningful VOCs were identified using minimum redundancy maximum relevance feature selection. To evaluate the distinguishing ability of these VOCs, a support vector machine was trained and validated using leave-one-out cross-validation.  
 
A cross-validation accuracy of 85.44% (sensitivity 77.42%, specificity 90.63%) was achieved using 26 unique VOCs (30 features). The selected VOCs included dimethyl sulfide, hexanol, 2-nonanone, methanol, and n-undecane, compounds previously indicated as possible biomarkers. These preliminary results show promise for a breath-based screening system using infrared CRDS and a spectral fitting program for biomarker identification.