Study On Detection System And Method Of Lung Cancer Related Markers Of Breathbased On Dual Signal

Lung cancer is a high risk of malignant tumor.Because of the lack of effectivetechnique for early diagnosis, its mortality rate has been the first among all malignanttumors.Some volatile organic gases (VOCs) in the exhaled air of human body areclosely related with whether they suffering from lung cancerand even the cancer degree.Thus these VOCs can be used as diagnostic markers of lung cancer. Compared with thetraditional detection methodsfor lung cancer, to diagnose on the basis of exhaled gas notonly lower cost, more importantly it is non-invasive, so it is expected to screening forearly lung cancer, which can greatly improve the survival rate of patients with lungcancer.At present, for the detection of lung cancer exhaled gas, there mainly areelectronic nose/tongue as emerging technologies, whichis based on sensor detectiontechnology, in addition to gas chromatograph, mass spectrometry and some otherlarge-scale instrument.Although the large-scale instrument is precision, it is tooexpensive for screening. However, in the current study, although the theory and processof electronic nose/tongue appear to perfect, the signalis sosingle that can only reflectspart of finite characteristic, and they are suitable for detection of a single substance,which is commonly used in industry.They are difficult to widely apply toactualscreening. In this study, based on theVL (visible light)-LIF (laser-inducedfluorescence)dual signal, a novel detection system and methodis developed forrapid,qualitative and quantitative detection of exhaled markers of lung cancer, which providesa new idea for the detection and screening for lung cancer. This paper mainlycompletedthe following distinctive works：1. Construction of a complete VL-LIF dual signal detection system of lung cancerexhaled breath markersStarting from the construction form of the sensor array, around the five keytechnical problems, the design of overall system was elaborated.The whole complexsystem was dividedinto four subsystems to construct one by one, includinggas pathsystem, optical system, mechanical system, and circuit system. A number of technicalproblems, design of reaction chamber, parameter setting of detection, form the lightsource, accurate positioning of movement, and connecting communication interface hadbeen effectively resolved through comprehensive utilization of modularization idea,theoretical model calculation, simulation experiment design, time division multiplexingtechnologyand other means. Finally, throughthe system engineering method, a completeVL-LIF dual signal detection system for lung cancer exhaled breath markers was madeup successfully by gettingthe four subsystemstogether organically, and the work flowwas established. The error analysis was carried out for the key parts, and the resultsshowed that it can meet the expected requirement. The construction of system had laid asolid material foundation for subsequent algorithm design and experimental study. Thissystem has applied for7patents, two of which have been authorized, while theremaining five havebeen public.2. Design of time-frequency multi-index feature vector extraction algorithmFrom the theories and concepts onthe feature, based on the analysis of importantcharacteristics and the mechanism of visible light signals and the laser-inducedfluorescence signals, through the pretreatment on the tow kinds of signals, thetime-frequency multi-index feature vector based on thedual signal was proposed byemploying gridding, threshold segmentation as image processing method and wavelettransform as mathematical tools. Multiple output forms of the feature vector wereobtained, including visualization map and fluorescent element coding and so on. Thedesign of corresponding algorithm provided the basis and laid the foundation for thesubsequent gas detedtion.3. Research and explore the unsupervised pattern recognition method based on themulti-index feature vector and test the recognition performance of the systemAccording to the characteristics of multi-index feature vector of double signalsystem, hierarchical cluster, neural network, nondimensionalization, principalcomponent analysis were used to qualitatively and quantitatively distinguish the targetgases. An unsupervised pattern recognition method was explored, based on themulti-index feature vector.Throughanalysing and exposing the success and failure ofhierarchical cluster for the different samples and the choice ofnondimensionalizationmethod in detail, the dual signal joint feature was proposed onthe basis of the analysis of the contribution of each signal for the recognition effect. Theresults shown thatthe qualitative identification rate of the system and the algorithm was100%,and the quantitative detection rate was up to97.4%based on a large number of repeated experiments.Finally, by means of uniting the system and the extraction andrecognition algorithms of the dual signal joint feature, a complete set of detectionmethod based on VL-LIFdual signal model for exhaled breath markers of lung cancerwas established.