Application Of SVM In Automatic Detection Of Minimal Residual Disease In Acute Myeloid Leukemia

Objective:To explore the use of support vector machine (SVM) algorithm to chieve an automated, objective, standardized analysis for acute myeloid leukemia(AML) minimal residual disease by flow cytometry analysis.Methods:Selected36cases expressing positive CD7and whose AML residual disease have been monitored more than once of all AML patients from2010to2012in our hospital, then exported the data of initial and residual disease review in turn. Using support vector machines, these two groups of datas were set as the training object, optimized parameters, established subject-independent predictive models. Finally this model was used for subject-independent residual disease automated analysis. Compared with the conventional manual analysis of residual disease data statistically, evaluated the reliability of automatic analysis. Results:SVM established and optimized subject-independently predictive models. Automated analysis of residual disease based on this model showed no significant statistical difference between conventional manual analysis. The correlation is significant (two-tailed paired t-test correlation coefficient0.986, p>0.05).Conclusions:SVM can assist flow cytometry for residual disease monitoring with automated, objective and standardized analysis.