NEURAL-NET IDENTIFICATION OF THUMB MOVEMENT USING SPECTRAL CHARACTERISTICS OF MAGNETIC CORTICAL RHYTHMS

Neural nets have shown great promise as tools for reducing and examini ng multi-dimensional data. When carefully tuned with selected data set s of individual subjects neural nets have indisputable potential in id entifying distinct stages of voluntary finger movements. However, robu st, automatized data description methods would be needed to eventually extend the use of neural networks into visualization of brain activit y during more complex, multimodal tasks where the cortical processes a re not equally well understood. We explored the suitability of a self- organizing map (SOM) in the widely studied case of voluntary finger mo vements (left and right thumb), using as input such spectral character istics that showed systematic task-dependent changes when averaged ove r repeated movements. SOMs constructed without individual fine-tuning and with generally chosen training parameters from these spectral feat ures identified correctly 85% of the ongoing movements but, somewhat s urprisingly, not the side of thumb movement. Even for this inclusive c hoice of input, the neural nets were sensitive to transient signals, b ut focused fine tuning, based on a priori known subgroups in the data, is clearly required for more detailed classification. Thus, a neural net visualization is likely not the most attractive first approach for characterization of cortical processing during complex multimodal tas ks.