Independent component analysis, a new concept?
Signal Processing - Special issue on higher order statistics
Self-organizing maps
A neural classifier enabling high-throughput topological analysis of lymphocytes in tissue sections
IEEE Transactions on Information Technology in Biomedicine
BSS-Based Feature Extraction for Skin Lesion Image Classification
ICA '09 Proceedings of the 8th International Conference on Independent Component Analysis and Signal Separation
Hi-index | 0.00 |
We consider the task of cell classification in fluorescent micrographs. We combine the use of Independent Component analysis as a preprocessing step and a Self-organizing Map for the resulting ICA feature space to classify image patches into cell and noncell images and to investigate the features of image patches in the vicinity of the classification border. We compare the classification performance of ICA bases of different size, generated from applying the infomax algorithm to image eigenspaces of different dimensionalities. We find an optimal performance for intermediate dimensionalities, characterized by the ICA basis patterns exhibiting salient features of an "idealized" cell shape, and we achieve classification results comparable to a previous approach based on PCA features.