Abstract: PURPOSE: Optical imaging based on intrinsic signals (OI) is widely used for the functional architecture studies of cat primary visual cortex. In such applications, a powerful tool to detect orientation columns is blind source separation (BSS) technique. Being data-driven methods, BSS approaches do not require any prior information. However, incorporating prior information such as experimental paradigm can be helpful for the accurate detection of activations. In our study, prior information is incorporated to the canonical correlation analysis (CCA), and an experimental paradigm information guided BSS method was proposed for the analysis of cat primary visual cortex OI data.
METHODS: The amplitude of the response signal change significantly after stimulus, therefore a mean difference before and after stimulus can be used as a temporal constraint to guide the separation of response signal. The proposed method is obtained by imposing the temporal constraint to the estimated time courses of CCA using experimental paradigm.
RESULTS: The proposed method is evaluated and validated by comparisons with the conventional method on cat primary visual cortex OI data. Comparing with the conventional method, our approach detected more accurate orientation columns and achieved more consistent performance among different trials. The results obtained by the proposed method is more reasonable and interpretable.
CONCLUSIONS: Incorporating experimental paradigm information into BSS can improve the accuracy and robustness of pure data-driven method. The proposed method is suitable for the analysis of cat visual cortex OI data and can obtain a more accurate and interpretable activation for the detection of orientation columns.

Key words: BSS, orientation column detection, primary visual cortex, prior information, OI