Mai

Introduction to retinotopy

Early visual areas, located in the posterior occipital lobe, are retinotopically organized: A particular location in the early visual cortex responds to stimulation at a particular location in the visual field, while neighboring locations on the cortical surface respond to neighboring locations in the visual field. This pattern of response maps a representation of the visual field onto the cortical surface.

Retinotopic areas on example subject

These early visual areas are mapped with two parameters: eccentricity and polar angle. Eccentricity indicates distance from fovea. The most posterior regions of the early visual areas has a pronounced foveal preference, while more anterior regions prefer more eccentric stimuli. In visualizations of retinotopy, eccentricity mapping appears as concentric "rings" of different colors, each denoting a different eccentricity. Polar angle indicates the angle from the horizontal meridian, so polar angle mapping shows the reversals in angular representations on the cortex, which are used in defining retinotopic maps.

While the primary visual cortex (V1) was the first visual field map discovered, over the last 15 years, a variety of retinotopic maps have been identified on the ventral and dorsal surface of visual cortex. While definitions of these visual areas will not be discussed, more information about these retinotopic maps can be found here. Information about defining these regions can be found here and here.

Traveling wave model

While retinotopic maps were first identified in the 1940s ago using electrophysiology, it is only in the last 20 years that these maps have been identified using fMRI. The turning point was in 1994 when Engel, Glover, and Wandell developed a new method for mapping visual field maps using a "traveling wave" of activation. First used to map eccentricity in V1, this method was a vast improvement on previous methods. In a nutshell, this method activates different regions of the visual cortex at different times, creating a "traveling wave" of activation from one region to another. The delay in activation allows us to infer the eccentricity or polar angle preference of a region. So how does this actually work?

The Stimuli

Stimuli used in mapping polar angle (top) and eccentricity (bottom)

Early visual cortex has both eccentricity and polar angle mapping, so two types of stimuli are needed. Eccentricity is mapped using a series of expanding, concentric checkerboard rings while polar angle is mapped by a rotating checkerboard wedge. These checkerboard patterns have been demonstrated to generate a strong neural response in early visual areas.

These stimuli are designed to activate adjacent retinotopic voxels sequentially, such that voxels with different eccentricity (or polar angle) preferences are activated at different times. For example, let's consider a hypothetical row of retinotopic voxels along the calcarine sulcus, from posterior to anterior:

We know the most posterior voxels (the two boxes furthest to the left) have a foveal preference. In an eccentricity experiment with a expanding series of rings, these voxels will therefore be activated really early on (t = 0, hypothetically). The two voxels in the middle are more anterior, so they probably have a more peripheral preference relative to the first two voxels, so they won't be activated until later in time. The most anterior voxels with the most peripheral preference will be activated last in time, so there is a "traveling wave" of activation from posterior to anterior (thus the name "traveling wave model").

This seems like a really simple idea, but it's absolutely critical to understanding the traveling wave model.

• More posterior voxels are activated first
• Most anterior voxels are activated later

Phase Encoding

The traveling wave of activation created by the stimuli means that we can extract (in this case) eccentricity preference from the time series for a retinotopic voxel.

Let's think about the hypothetical time series for a hypothetical retinotopic voxel:

This time series has 6 peaks because a typical retinotopic scan has 6 cycles of the stimulus. More importantly, we see that the signal starts rising at t = 0. This hypothetical voxel must have been activated very early in the cycle, which means that it was probably activated by a foveal stimulus--so this voxel probably has a foveal preference.

In contrast, let's look at another hypothetical time series, that is almost identical to the one above:

Like the first time series, this one also has 6 peaks--but there's a delay before the signal begins rising. It is "phase-shifted" to the right, which just means that this voxel was activated later in time relative to the first one. If it was activated later in time, it was probably activated by a more peripheral stimulus, so this voxel probably has a peripheral preference. Thus, voxels activated later in time are "phase shifted," so the location preference for a voxel is "phase encoded." Given a voxel time series, we can find its eccentricity (or polar angle) preference by determining its phase.

Finding the phase

raw time series for example voxel

In practice of course, the time series for a retinotopic voxel never looks that nice. To the right is the unfiltered time series for a real voxel. While the six peaks corresponding to the six cycles of the experiment are readily apparent, there is still a significant amount of noise from signal drift, thermal noise, and physiological noise. This scan has already been motion corrected.

filtered time series

Thus, before this voxel's phase can be determined, the time series needs to be filtered to remove noise. Here, the time series was detrended by removing the baseline and then dividing by the mean signal. The resulting filtered time series still has the six peaks, but the most obvious noise has been removed so that the peaks are varying around 0. Now we can think about finding the phase, and therefore location preference.