David M. Senseman and Kay A. Robbins
University of Texas at San Antonio
Under Submission
Abstract
In the pond turtle, Pseudemys scripta elegans, visually evoked cortical waves propagate at different velocities within the primary visual area compared to waves that pass into the secondary visual area. In an effort to separate intra- and inter- cortical wave motions, movies of visually evoked cortical waves recorded by high speed voltage-sensitive dye (VSD) imaging were subjected to Karhunen-Loeve (KL) decomposition. This procedure decomposes the VSD movies into a series of basis images that capture different spatial patterns of coherent activity. Most of the energy of the compound wave motion (> 95%) was captured by the three largest basis images, M1,1, M1,2 and M2,1. Based on visual comparison with maps of wave front latency, KL basis image M1,2 appears to capture the spread of depolarization within the primary visual area while KL basis image M2,1 appears to capture the spread of depolarization from the primary into the secondary visual area. The contribution of different basis images to the intra- and inter- cortical wave motions was tested by reconstructing the response using different combinations of KL basis images. Only KL basis images M1,1 and M1,2 were needed to reconstruct intra-cortical wave motion while basis images M1,1 and M2,1 were needed to reconstruct inter-cortical wave motion. It was also found that the direction and speed of wave propagation could be deduced by visual inspection of the basis image projections on to the original data set. The relative advantage of KL decomposition for the analysis of complex wave motions captured by VSD imaging is discussed.Key words: cerebral cortex; visual cortex; neural network; voltage-sensitive dye; turtle vision, KL decomposition, principal component analysis, cortical model
The supporting materials provided with this paper include 6 movies that are referred to in the text plus visualizations of the spatial and temporal modes for all of the trials in the two preparations (Cor34: Left Cortex and Cor32: Right Cortex) that are used in this paper. The actual movies for this and othe data sets referenced in the paper can be found at http://vip.cs.utsa.edu/cortex/Papers/CorticalWaves. Alternatively these materials are available on CD-Rom and will be provided at no cost by the authors.
Movie #1: A side-by-side comparison of the original data with a data set reconstruction using only the three largest modes (M1,1, M1,2 and M2,1 for this data set). Individual frames of this movie are displayed in the left two columns of Fig. 5. The reconstructed movie captures the prominent characteristics of the rostrocaudal and dorsomedial propagation--namely the constant velocity of the cortical wave during its propagation along the rostrocaudal axis and its pronounced slowing as it travels dorsomedially into DM. Trials cor3409 and cor3207 are displayed.
Movie #2: A movie reconstructed using only mode M1,1, corresponding to the top row of Fig. 6Aa. The reconstructed data shows a lock-step rise and fall. Because of the pseudo-color scheme, the visual appearance is dominated by the location of the peak of M1,1, making it appear so though there is a source located centrally in VC. Trials cor3409 and cor3207 are displayed.
Movie #3: A movie reconstructed using only modes M1,1 and M1,2, corresponding to Fig. 6Ab. The movie partially captures the rostrocaudal movement. Trials cor3409 and cor3207 are displayed.
Movie #4: A movie reconstructed using only modes M1,1 and M2,1 corresponding to Fig. 6Ac. The movie partially captures the movement into DM. Trials cor3409 and cor3207 are displayed.
Movie #5: A movie animating the schematic diagram of Fig. 6Bb showing how two standing waves can add to make a traveling wave.
Movie #6: A movie of trial cor3207 corresponding to Fig. 7Bb, a trial with a secondary rise, but and a reversal of sign of the projection of M1,2. This trial has a wave reflection in the caudal pole. The color scheme used in this movie emphasizes the amplitude and latency of the wave front. Each pixel is displayed in neutral blue until its amplitude reaches half of its maximum height. At this time the color of the pixel is changed, using a rainbow color map that reflects latency. Dark red corresponds to lowest latency, while deep purple corresponds to longest latency. Each time the pixel falls below its half-height, its color is reset to neutral blue. This color scheme shows propagation more clearly than a straight amplitude scaling.
Preparation Cor 34: The left cortex preparation used in Fig. 3, 4, 5, 8 and 9.
Preparation Cor 32: The right cortex preparation used in Fig. 3, 4, 5, 6, 7, 8 and 9.