EN
The examination of oligoclonal bands (OCB) in cerebrospinal fl uid (CSF) is important for the diagnosis of multiple sclerosis, however the correlation between quality, number and disease progression is uncertain. The aim of this study was to test the automatic system, for detection and identifi cation of OCB. The patterns of OCB, obtained from isoelectrofocusing of CSF proteins, were scanned with the use of fl atbed scanner. Next, for each of the scanned images, the selected features of the pattern were extracted by the means of the image processing algorithms, and arranged into the feature vector. That created the ìsignatureî of the image that was subsequently analyzed by classifi er based on the Artifi cial Neural Networksí. The result was the positive (P), negative (N) or ìuncertainî (U) classifi cation of the bandsí pattern. We have used database of the 225 samples, manually classifi ed by the expert, that formed the training set for the classifi er with the equal number of positive and negative results. Among 20 samples (10P/10N) used in the testing phase of the system 17 were classifi ed correctly, 3 were ìuncertainî, no false results was obtained. The system is implemented in the MATLAB environment. The future work focus on designing the system that would be able not only to classify the OCB patterns, but also to possibly cluster the images into the groups with common parameters.