Analysis Of High Dimentional Mass Cytometry Data Based On Support Vector Machine And Its Application In The Early Diagnosis Of Acute Myelocytic Leukemia

Leukemia is a malignant tumor in hematopoietic system. It begins in bone marrow and results in the malignant proliferation of abnormal white blood cells. Leukemia ranks as the 6th common cancer in china. It is the top ranked cancer in teenagers. According to the speed of onset of the disease, Leukemia can be classified into two categories:acute leukemia and chronic leukemia. Acute myeloid leukemia (AML) is a cancer of myeloid blood cells, and it is the most common form of adult acute leukemia.The clinical diagnosis methods of acute myeloid leukemia include blood tests, bone marrow routine examination, cellular immune typing and cytogenetic examination. These methods are not automated analysis and require experienced clinical pathologists. Besides, they are time-consuming, subjective and have other limitations. In clinical, AML is diagnosed when blood or bone marrow myeloblasts (or monocytes) cells are above 20%. The automation of early diagnosis of AML is important because it can improve the diagnostic accuracy and patient cure rate.Mass cytometry is a new type of single-cell analysis technology developed in recent years. The technology integrates the principles of mass and cytometer. In a single cell, dozens or even hundreds of feature markers can be measured. It is suitable for high-speed analysis. Also, it has high precision and high recognition accuracy. Compared with the traditional fluorescence flow cytometry, mass cytometry has great advantage in taking multi-parametric measurement. Since there is no interference between channels, it does not require compensation calculation. Consequently, mass cytometry is a new trend of single-cell analysis.In this thesis we first describe related theories of mass cytometry, including its current development, principles, data analysis methods commonly used and application. Then we introduce a data analysis method based on support vector machine (SVM), which is built specifically for analysizing the high-dimentional data from mass cytometry. In the thesis we use this platform analyzing mass cytometry data taken from healthy human bone marrow samples. We successfully multi-classified the bone marrow cells and then assessed the performance of the platform with the classification results of data analysis. In the last part of the thesis, we carry out research on diagnosis of the early stage of AML. We preliminarily verified the feasibility of this method for automation diagnosis of AML at the early stage.Mass cytometry can measure a great number of parameters at the same time with high single cell recognition accuracy. It is expected to obtain better diagnostic accuracy with broad prospect. This thesis combines the advantages of high recognition accuracy mass cytometry and the machine learning method of SVM, reveals its potential in increasing the early diagnosis level of AML. Our work has a great significance in early diagnosis and automating the process.