Automatic lineage assignment of acute leukemias by flow cytometry

A method for automatic lineage assignment of acute leukemias was developed. Input are eight list mode data files acquired with a FACScan flow cytometer. For each cell, four parameters are measured: forward light scatter, orthogonal light scatter, fluorescein fluorescence, and phycoerythrin fluorescence. Eight data files are acquired in the following sequence: unstained, isotype controls, CD10/CD19, CD20/CD5, CD3/CD22, CD7/CD33, HLADR/CD13, and CD34/CD38. First, each of the data files 3 to 8 are clustered independently employing an algorithm based on nearest neighbors. Next, the clusters are associated across the data files to form cell populations, using the assumption of light scatter invariance across tubes for each population. The mean positions of each cell population are fed into a decision tree. The decision tree first identifies normal cell populations, i.e., monocytes, neutrophils, eosinophils, basophils, NK cells, T-lymphocytes, and B-lymphocytes. After elimination of the normal cell populations from the data space, the residual cell populations are classified as B-lineage ALL, T-lineage ALL, AML, AUL, B-CLL, or unknown. The effectiveness of this novel approach is shown with case studies of B-lymphoid, T-lymphoid, and Myeloid acute leukemias.