Neural Network Analysis of Flow Cytometric Data

Multidimensional flow cytometry allows the discrimination of neoplastic cells from their normal counterparts. Using technology for the simultaneous measurement of 5 parameters, the abnormal expression of normal antigens can be used to identify leukemic populations. Distinction between normal and abnormal cells can be made by a highly trained scientist using a computer program which permits visualization of multidimensional data. We will demonstrate the utility of artificial neural network technology to reproduce an identification made by a trained scientist. We will focus on two forms of acute leukemia, acute lymphoblastic leukemia (ALL) and acute myeloblastic leukemia (AML), as test specimens to be compared with normal bone marrow specimens. Using this technology, we will characterize acute leukemias and report their relative positions in a multidimensional data space. We will also investigate whether the neural network can extract more information from the data set than the expert scientist.