Automated analysis of acute myeloid leukemia minimal residual disease using a support vector machine

// Wanmao Ni 1 , Beili Hu 2 , Cuiping Zheng 3 , Yin Tong 4 , Lei Wang 1 , Qing-qing Li 1 , Xiangmin Tong 1 , Yong Han 1 1 Key Laboratory of Tumor Molecular Diagnosis and Individualized Medicine of Zhejiang Province, Zhejiang Provincial People’s Hospital, Hangzhou, Zhejiang, P. R. China 2 Medical of College, Zhejiang University, Hangzhou, Zhejiang, P. R. China 3 The Center Hospital of Wenzhou, Wenzhou, Zhejiang, P. R. China 4 Department of Hematology, Shanghai General Hospital, Shanghai, P. R. China Correspondence to: Xiangmin Tong, email: tongxiangmin@163.com Yong Han, email: onlune@126.com Keywords: flow cytometry, immunophenotyping, support vector machine, acute myeloid leukemia, minimal residual disease Received: July 04, 2016      Accepted: September 29, 2016      Published: October 04, 2016 ABSTRACT We investigated the ability of support vector machines (SVM) to analyze minimal residual disease (MRD) in flow cytometry data from patients with acute myeloid leukemia (AML) automatically, objectively and standardly. The initial disease data and MRD review data in the form of 159 flow cytometry standard 3.0 files from 36 CD7-positive AML patients in whom MRD was detected more than once were exported. SVM was used for training with setting the initial disease data to 1 as the flag and setting 15 healthy persons to set 0 as the flag. Based on the two training groups, parameters were optimized, and a predictive model was built to analyze MRD data from each patient. The automated analysis results from the SVM model were compared to those obtained through conventional analysis to determine reliability. Automated analysis results based on the model did not differ from and were correlated with results obtained through conventional analysis (correlation coefficient c = 0.986, P > 0.05). Thus the SVM model could potentially be used to analyze flow cytometry-based AML MRD data automatically, objectively, and in a standardized manner.