Analysis of biomédical textured images with application of synchronized oscillator-based CNN

This paper is focused on the analysis of biomedical images, including textured ones. A segmentation method, based on network of synchronized oscillators is presented. Oscillator networks can be considered as a special case of the CNN. Its oscillatory dynamics allows encoding the different features of objects forming the visual scene, thus makes these network suitable for medium level image processing, like image segmentation. Oscillator networks can process both two and three dimensional images. The proposed method was tested on several biomedical images acquired with the use of different modalities. Principles of operation of the oscillator networks are described and discussed. Obtained segmentation results for sample 2D and 3D biomedical images are presented and compared to image segmentation based on multilayer perceptron network (MLP).