Ds. Frankel et al., APPLICATION OF NEURAL NETWORKS TO FLOW-CYTOMETRY DATA-ANALYSIS AND REAL-TIME CELL CLASSIFICATION, Cytometry, 23(4), 1996, pp. 290-302
Conventional analysis of flow cytometric data requires that population
identification be performed graphically after a sample has been run u
sing two-parameter scatter plots, As more parameters are measured, the
number of possible two-parameter plots increases geometrically, makin
g data analysis increasingly cumbersome. Artificial Neural Systems (AN
S), also known as neural networks, are a powerful and convenient metho
d for overcoming this data bottleneck, ANS ''learn'' to make classific
ations using all of the measured parameters simultaneously, Mathematic
al models and programming expertise are not required, ANS are inherent
ly parallel so that high processing speed can be achieved. Because ANS
ate nonlinear, curved class boundaries and other nonlinearities can e
merge naturally, Here, we present biomedical and oceanographic data to
demonstrate the useful properties of neural networks for processing a
nd analyzing now cytometry data, We show that ANS are equally useful f
or human leukocytes and marine plankton data, They can easily accommod
ate nonlinear variations in data, detect subtle changes in measurement
s, interpolate and classify cells they were not trained on, and analyz
e multiparameter cell data in real time, Real-time classification of a
mixture of six cyanobacteria strains was achieved with an average acc
uracy of 98%. (C) 1996 Wiley-Liss, Inc.