用于GC-IMS的呼气分析数据处理方法OA

A novel data processing method of gas chromatography-ion mobility spectrometry(GC-IMS)is proposed to address the challenges of peak overlapping in GC-IMS spectra for exhaled breath analysis and the limited and continuously updated samples in applications.Breath samples simulating different physiological states were collected from healthy volunteers before and after drinking coffee,and were ana-lyzed by GC-IMS directly.Overlapping peaks in the GC-IMS two-dimensional spectra were resolved using a Gaussian second derivative peak sharpening algorithm,and a Mondrian forest(MF)incremental learning model was constructed for classification of the two states.The results indicate that the method successfully resolves overlapping peaks in the original GC-IMS spectra,increasing the peak signal-to-noise ratio(PSNR)to 50 dB.Features such as the positions and intensities of the resolved peaks in the GC-IMS spectra are extracted to build the MF incremental learning model,which maintain a high average classification accuracy of 93.75%as samples increase,significantly outperforming comparative models like random forest and Hoeffding trees.This exhaled breath analysis data processing method enhances classifi-cation accuracy,showing promising practical application prospects.