Mid-infrared acetone sensor for exhaled gas using FWA-LSSVM and empirical mode decomposition algorithm

Abstract

The acetone can be regarded as one of the respiratory biomarkers for lung cancer, the detection of acetone in human exhaled gas has great important for early lung cancer screening. Based on Tunable Diode Laser Absorption Spectroscopy (TDLAS), a system for detecting acetone in exhaled gas has been developed, a multi-pass gas cell (MPGC), which has a 76 m effective absorption optical path, as well as a 7.32 μm quantum cascade laser (QCL) were adopted in the detection system. In order to confirm the function of the system, a gas distribution station equipped with standard concentrations of acetone was used. Empirical mode decomposition (EMD) algorithm as a spectral pre-processing algorithm was implemented to reduce the noise from the second harmonic signal (2f signal), using an orthogonal lock-in amplifier based on LabVIEW to demodulate the detector signal. Fireworks algorithm with the least squares support vector machine (FWA-LSSVM) was designed to improve the accuracy and sensitivity of acetone gas concentration inversion. The limits of detection (LoD) of the sensor system can be found the middle value of up to ∼ 21.4 parts per billion (ppb) with a 78 s integration time by Allan-Werle deviation analysis. The exhaled breath experiment showed that the sensor provides a promising way for the detection of acetone in human exhaled breath.