Three applications of pulse-coupled neural networks and an optoelectronic hardware implementation

摘要

Abstract: Pulse Coupled Neural Networks have been extended andmodified to suit image segmentation applications.Previous research demonstrated the ability of a PCNN toignore noisy variations in intensity and small spatialdiscontinuities in images that prove beneficial toimage segmentation and image smoothing. This paperdescribes four research and development projects thatrelate to PCNN segmentation - three different signalprocessing applications and a CMOS integrated circuitimplementation. The software for the diagnosis ofPulmonary Embolism from VQ lung scans uses PCNN insingle burst mode for segmenting perfusion andventilation images. The second project is attempting todetect ischemia by comparing 3D SPECT images of theheart obtained during stress and rest conditions,respectively. The third application is a space scienceproject which deals with the study of global auroraimages obtained from UV Imager. The paper alsodescribes the hardware implementation of PCNN algorithmas an electro-optical chip. !5