The Experimental Study Of Laser Induced Breakdown Spectroscopy On The Diagnosis Of Human Cancer

In clinical medicine,rapid and accurate diagnosis of cancer is an important subject.The method of existing pathological diagnosis has many shortcomings,such as complicated pretreatment of tumor samples,long time consuming and strong dependence on the experience of the doctor.Owing to the advantages including simple devices,without pretreatment of samples and online detection,laser-induced breakdown spectroscopy(LIBS)has become a powerful tool for cancer diagnosis.Now LIBS has been applied to the classification and identification of tumor tissues,but the diagnosis method based on LIBS of human blood has not yet been researched.Based on this fact,the dissertation will carry out the research on human cancer diagnosis by using LIBS of whole blood and serum.Firstly,the dissertation introduces the research progresses of LIBS in biological sample detection,such as qualitative or quantitative detection of chemical elements,classification and identification of biological tissues.By analyzing the present situation in cancer diagnosis,the dissertation presents the main research contents of this paper.Later,it introduces the mechanism of LIBS and gives the methods which are principal component analysis(PCA),k-nearest neighbor(k NN),support vector machine(SVM).Utilizing these algorithms,the dissertation builds the mathematical model based on LIBS of human whole blood and serum to achieve the diagnosis of human cancer.Moreover,it provides cross validation and confusion matrix by which to evaluate classification model and the performance of the diagnosis.Secondly,using frequency 5Hz,1064 nm Nd:YAG laser ablation whole blood samples and serum samples which is fresh or melted.These samples were collected from patients who are suffering from cancer of lymphoma,leukemia,multiple myeloma,stomach,liver or pancreatic and normal human individuals.By analyzing the LIBS of these samples,the dissertation establishes a database including spectral lines irradiating by the atoms of Fe,Ca,Cu,Na,Mg,Al,N,K,O,H etc.Moreover,by utilizing PCA algorithm,it shows that a significant difference between LIBS of blood samples from patients and normal people.The difference also existing between LIBS of blood samples from patients suffering from different cancers.Finally,utilizing the LIBS data of blood samples,weighted-k NN classifiers and SVM classifiers with three order polynomial kernel function are established.These models are trained and evaluated by 10-fold cross validation and confusion matrix.The results show that by using fresh whole blood samples the diagnostic accuracy of cancer can reach about 100% and the diagnostic accuracy for cancer of leukemia,lymphoma,and multiple myeloma are about 88%,96% and 92% respectively.The diagnostic accuracy of cancer can be about 100% with melted whole blood,and the diagnostic accuracy for cancer of leukemia,lymphoma and multiple myeloma are about 88%,95% and 89% respectively.By using fresh serum,the diagnostic accuracy is about 100%,and for cancer of liver,lymphoma,multiple myeloma,pancreatic and stomach the diagnostic accuracy are about 87%,98%,85%,75% and 82% respectively.Finally,using melted serum,the diagnostic accuracy is about 100%,and for cancer of lymphoma,multiple myeloma,pancreatic and stomach,the diagnostic accuracy are about 68%,81%,29%,90%,and 84% respectively.At last,these results indicated that whole blood samples is performing better than serum samples for the purpose of cancer diagnosis,and the fresh samples are more suitable than melted samples for cancer diagnosis.The research in this paper has huge application prospect in the issue of rapid and accurate diagnosis of cancer.Moreover,it provides valuable proposals for the classification of biological tissues based on LIBS.