Abstract:In this paper, a novel structured noises-reduction technique for OI data is proposed. Canonical correlation analysis (CCA) technique is exploited to separate the underlying independent sources among which the neural response signal is picked out by the correlation analysis. The white noise (WN) criterion is applied to discern the structured components from the unstructured ones. The energy of structured noises is then eliminated from the original data. Monte Carlo simulation is used to test the validity of the procedure. The result shows that after the noise reduction, the true positive rate improves significantly without raising the false positive rate. Five sets of OI data of single trial collected from the HP area of rat's cortex are processed by the procedure and the resulting activation maps present more detailed spatial architecture than those without noise reduction.

Structured Noise Analysis in Intrinsic Optical Imaging