Study On Auxiliary Detection Methods Of Lung Cancer And Air Pollution Factors Affecting Lung Cancer

Lung cancer threatens human health seriously,and its morbidity and mortality rank first in all cancers.The high incidence and mortality of lung cancer are affected by many factors.Firstly,as a morphological lesion detection method,CT is difficult to detect early lung cancer in the functional lesion stage.Electrical Impedance Tomography（EIT）,as a new functional imaging method,has very good prospects in early lung cancer detecting,but its resolution is far lower than the clinical requirements.Therefore,at present,the detection rate of early lung cancer is low,and most patients with lung cancer are in the middle and late stages of treatment,thus losing the best opportunity for treatment.Secondly,the mainstream CT technology used in lung cancer diagnosis has limited recognition rate and low intelligence level.A large number of CT images are checked by doctors,which results in heavy workload.Thus,the fatigue easily lead to misdiagnosis and missed diagnosis.Thirdly,air pollutants have become one of the main pathogenic factors of lung cancer,which have a significant impact on lung cancer incidence.Based on the above background,this paper deeply studied the electrical characteristics of lung cancer tissues,the EIT detection of lung cancer,computer-aided detection of lung nodules based on CT images,and the effects of air pollutions on lung cancer incidences.Also,the correlation between CT image characteristics and conductivities of lung cancer tissues was preliminarily explored.The details and results are as follows:1.The impedance spectra of human active lung cancer tissues and normal lung tissues were measured.And the impedance spectrum model,impedance and dielectric spectrum characteristics of lung cancer and normal tissues were studied.Then,the conductivity distribution models of human lung were established to study the distribution rule and change rule of conductivity.The analysis results showed that the conductivity of the left lung was higher than that of the right lung,and the conductivity of the different lobes was also different,showing certain regularity.Lung cancer could cause the increase of conductivity,and the increase of conductivity of left lung（34.6%）was more obvious than that of right lung（31.4%）.The conductivity of the posterior base of left lower lobe and the apical area of right upper lobe changed significantly.Overall,conductivity increased 33.5%on average.Based on the conductivity distribution characteristics of human lung,the Tikhonov regularized algorithm based on prior information（PI-TR）was proposed,and a mathematical method of solving regularized parameter based on phase holdup was proposed.Comparative simulation experiments of lung cancer detection by EIT were carried out.The results show that compared with the existing Tikhonov regularization（TR）algorithm,PI-TR can reduce the relative error of EIT lung cancer detection reconstruction image by 47.1%on average,and increase the correlation by 35.3%on average;compared with the existing L-curve method,the proposed method of solving regularized parameters can reduce the relative error of TR and PI-TR reconstruction image by 9.2%on average,and increase the correlation by 29.9%on average.2.An unsupervised clustering validity index was proposed,to estimate the optimal number of clusters in a dataset without any clustering algorithm.It effectively overcomed the limitations that the existing validity indexes were supervised and their robustness was poor.Further,the fuzzy c-means clustering algorithm based on MSO index,XB index and PC index was used to detect pulmonary nodules,respectively.The three methods were respectively recorded as M-FCM,X-FCM and P-FCM.The results showed that compared with X-FCM,M-FCM improved the accuracy of pulmonary nodules detection by an average of 3.3%,and the detection efficiency by an average of 80.9%;compared with P-FCM,M-FCM improved the accuracy of pulmonary nodules detection by an average of 7.35%,and the detection efficiency by an average of 74.6%.The above results verified that MSO index can effectively improve the accuracy and efficiency of lung nodule detection by clustering algorithm,and can better assist doctors to interpret CT images and detect lung nodules.3.Choquet integral association analysis method based on extended?-fuzzy measure was proposed.This method can not only evaluate the effect of typical air pollutant on different types of lung cancer and different sex groups,but also evaluate the interaction between different pollutants.And a rule mining method based on rough set was proposed.Based on the actual data samples,the association rules between air pollutants and lung cancer incidences were objectively excavated.Taking the air pollutants and lung cancer incidences data in Tianjin as an example,the analysis results show that SO₂,O₃,PM_2.5 and NO₂ are the major factors in fluencing lung cancer incidences;SO₂,NO₂ and CO have obvious effects on squamous cell carcinoma,while O₃,PM_2.5 and PM₁₀have prominent effects on adenocarcinoma;men are more sensitive to SO₂,while women are more sensitive to PM₁₀;there are obvious positive interaction between O₃ and PM_2.5,and among SO₂,PM_2.5 and PM₁₀.There are negative interactions among SO₂,O₃ and PM₁₀,SO₂,NO₂and PM₁₀,and SO₂,NO₂ and CO.The degree of interaction is different for adenocarcinoma and squamous cell carcinoma.4.The characterization of CT image features of adenocarcinoma tissues by conductivity was preliminarily explored.According to correlation coefficient and regression analysis,the relationship models between the standard deviation of inertia moment（texture feature of CT image）and conductivity of adenocarcinoma tissues were finally obtained,which reveals the correlation between CT image features and conductivity of adenocarcinoma tissues on a certain extent,and provides a basis for studying the conductivity characteristics of adenocarcinoma tissues from CT image characteristics.