Regensburg 2000 – scientific programme

Parts | Days | Selection | Search | Downloads | Help

DY: Dynamik und Statistische Physik

DY 46: POSTER II

DY 46.64: Poster

Thursday, March 30, 2000, 15:00–18:00, D

Analyzing flowcytometric fluorescence signals with neural nets – classification of fluorescence labeled blood cells with self-organizing neural nets — •O. Jäger, Ch. Ziegaus, Ch. Bauer, and E.W. Lang — Institut für Biophysik und physikalische Biochemie, Universität Regensburg, 93040 Regensburg

Artificial neural networks are well suited for feature extraction and classification of high-dimensional biomedical signals. Self-organizing feature maps (SOM) provide algorithms with unsupervised competitive learning rules, which allow to extract low-dimensional features from high-dimensional input data. SOM provides a topological mapping of similar input patterns onto nearby neurons, thereby finding inherent clusters in the low-dimensional feature space. In this investigation high-dimensional flowcytometric fluorescence signals of blood cells labeled with up to 7 different fluorescence markers were analyzed. Self-organizing map (SOM), dynamic cell structure (DCS), growing neural gas (GNG) and generative topographic mapping (GTM) algorithms have been used in a comparative study. Inherent clusters could be well isolated in certain projections and classified according to a number of prespecified classes.