A better result was obtained by assigning eccentricity according to the volume magnification function derived by Malpeli et al. (1996, J. Comp. Neurol. 375: 363-377; their Equation 2). The first step was to identify the projection column of the center of the fovea. The original data for the central 1° consists of 27 sites recorded along three microelectrode passes. The surface representing the horizontal meridian is fairly flat and well-behaved (approximating the natural plane of symmetry of the LGN), and one microelectrode pass (17 sites) intersected this surface at the exact center of the fovea, so the reconstruction of the horizonal meridian through the center of the fovea should be fairly accurate. The path of the foveola's projection column within the horizontal meridian representation was specified by identifying voxels at the points of convergence of sectors and setting their eccentricity to zero (FOVEOLA.DAT).
The central 1° of visual space was then divided into zones, each spanning 0.1° of eccentricity, and the volume of LGN (laminar space only) representing each zone determined by integrating the volume magnification function. Voxels were rank-ordered by eccentricity using a linear interpolation between the foveola and the 1° isoeccentricity surface to define a family of nested, smoothly changing isoeccentricity surfaces, down to voxel resolution. Voxels were then assigned eccentricities working from the foveola outward, rounding to the nearest 0.1°, according to the volume magnification function. For example, 0° eccentricity was assigned to voxels beginning at the lowest eccentricity and working toward higher eccentricity until their cumulative volume equaled the calculated volume of the central 0.05°. Then voxels from 0.05° onward were assigned the value 0.1° until their cumulative volume equaled the calculated volume of the zone 0.05° to 0.15°. This process was extended zone by zone, to the 1° isoeccentricity surface.
Last modified: February 19, 1999