Brain unfolding and flattening to view multiple visual areas

We have developed algorithms for stripping off the skull and then reconstructing the complete folded cortical surface of each cerebral hemisphere from structural MRI images. Once we have obtained a surface free of topological defects, it can be unfolded to better view the location of visual areas which in real life are partly hidden in deep fissures (sulci). This procedure works by gradually reducing curvature while trying to minimize areal and angular distortion.

Here's how the skull is first removed using a stiff deformable template.
The original cortical unfolding process descibed in Dale and Sereno (1993) is shown in the next movie (red => sulci, green => gyri).
In the next movie, one hemisphere is initially shown in ventral view. It then begins to unfold, while at the same time slowly rotating to a medial view. The visual areas are shown in yellow and blue (yellow => mirror-image, and blue => non-mirror-image visual field representation) while the position of the sulci are shown in grays (dark gray => sulci, light gray => gyri).
In the next movie, the same hemisphere is shown initially in a medial view. Once again it unfolds, but this time, slowly rotating to a posterior view.
In the next movie, the process by which portions of the cortex are completely flattened to a plane is shown. The posterior portion of the cortex containing retinotopically organized visual areas has been detached, and then cut a second time along the fundus of the calcarine sulcus. It was then flattened at once to a plane. The movie shows how the flattened surface is then unfurled on the plane by reducing areal and angular distortion. This process opens up the cut along the fundus of the calcarine which is visible at the left. In this case, the cortex has been colored just by curvature (red => sulci, green => gyri).
The next movie (same color code as previous) shows the flattening of the entire cortex. In this case, a series of cuts were made along the medial wall and the entire cortex allowed to unfurl on the plane.
The next movie shows the same full-cortex-unfolding sequence, except this time, distortion measures are plotted instead of curvature. Areal distortion is shown in color (red => expansion, green => contraction) and shearing distortion is shown as the length of small white line segments (a little difficult to make out at this magnification). There is an initial expansion (red) of the lateral surface which eventually resolves to a slight compression (green). The shearing distortion (white) is quite obvious around the circumference of the brain early in the unfolding.
The next movies show animations of the response to a rotating semicircle on a unfolded and flattened cortex in two subjects. The white regions in each frame represent the maximum response of the cortex at a particular phase of the rotating semicircular stimulus. The phase is also color coded (red is lower field, blue is horizontal meridian, and green is upper field).
The next movie shows a similar animation for the expanding ring stimulus. The pattern here is considerably simpler since isopolar angle lines run approximately perpendicular to the areal boundaries.
This movie shows an animation of the response to a bandpass noise stimulus where the center of the bandpass was swept from high to low frequencies. The peak responses are drawn on a representation of the gray/white matter boundary and the brain is tilted to show the superior temporal plane.
Same as last movie but shown on the unfolded cortical surface.
The next movie shows the result of deforming the inflated brain into an ellipsoid, which makes it possible to establish a spherical polar coordinate system (latitude and longitude) on the brain. The brain is rotated toward a medial view as it is deformed.
Once established on the ellipsoid, the coordinate system can be drawn on any surface. This movie shows the coordinate system drawn on the cortex as it unfolds into an area-corrected ellipsoid.
This movie shows how one ellipsoidal brain surface can be mapped to a second ellipsoidal brain surface in a point-to-point fashion using information about the curvature of the two brains (the target brain surface is not shown). By repeated application of this algorithm in a binary tree, a canonical brain surface can be constructed. A single point on this surface then indexes single points on all the surfaces from which it was made.
The movie shows the result of a small bug in one program for deforming the cortical surface...

Not a human, but quite folded.
For details, see Dale and Sereno (1993), Sereno et al. (1995), and Sereno et al., Dale et al., Tootell et al., Talavage et al., Krubitzer et al. forthcoming.

Back to the sereno homepage

Back to the cogsci homepage