Abstract

Optical Mapping (OM) sequences represent a very useful tool for registering cardiac bioelectric activity with high patio-temporal resolution. Conventional preprocessing of OM movies is mostly done with spatio-temporal filters, which could be improved for yielding better signal to noise ratio in the processed images. Also, the multimodal mapping, such as electrical mapping combined with OM, is limited by the need of presence of catheters in the optical image. In this work, we addressed a twofold objective. First, we explored the suitability and scope of intrinsic multi-recording techniques to separate different noise sources in OM sequences. Second, we explored the use of advanced techniques to extrapolate the electrical mapping shadowed spatial regions on the OM images. We scrutinized the Singular Value Decomposition (SVD) on OM sequences obtained in OM laboratory, showing that: (1) the second spatial component mostly conveys movement artifacts, and could be removed or processed for improving the quality; (2) the components from 3 to about 10 convey both spatial variations and spatial artifacts, which can not be readily decoupled. The use of GPatt, an advanced algorithm for multidimensional pattern extrapolation based on Gaussian Processes, was also analyze in a multimodal recording of OM and electrical mapping including a lasso catheter, showing interesting capabilities for removing the catheter shadows with fidelity, and dependence with respect to noise and texture. Our results show that other processing methods can provide with enhanced signal to noise ratio when working with OM sequences.