Optical Imaging of the Mammalian Visual Cortex

In vivo optical imaging of the visual cortex (both of intrinsic signals as well as with voltagesensitive dyes) makes the activity of entire neuronal assemblies accessible for the rst time. However, the magnitude of the data collected (>1 Gbyte/minute) as well as the tiny signal-tonoise ratio (O(10 )) necessitate the development of optimized analysis algorithms. We have developed extensions of the Karhunen-Lo eve principal components analysis that e ectively lter out unwanted noise and isolate the visually-relevant cortical response; the size of the remaining data set is drastically reduced. The implementation of these techniques on a Parallel Virtual Machine, which allows real-time extraction of cortical functional architecture and dynamics, is described as well. These techniques can be a powerful tool wherever manipulation of a large scienti c data base is required.