As was the case for exemplar identification, the accuracies of the category identification using voxels from only a single anatomical region were high; in some cases, these approached the accuracy obtained when the whole cortex was used (0.93 for left IES cortex, 0.83 for left SES cortex, and 0.82 for LIPL, vs. 0.98 for the whole cortex, for one of the participants).
Other brain areas also carry reliable information about individual tools and dwellings, demonstrating that the exemplar identification can be based on the neural representations of higher-level facets of the object properties.
For example, for one participant whose object exemplar identification accuracy based on the whole cortex was 0.94, the single-region accuracy was 0.77 for left superior extrastriate (SES), 0.77 for LIPL, and 0.82 for left inferior extrastriate cortex (IES).
The locations of voxels that underpinned this accurate object exemplar identification (i.e., the diagnostic voxels), were similar (at a gyral level) across participants, and were distributed across the cortex (as shown in Figure 3).
The highest exemplar identification rank accuracy obtained in this leave-one-participant-out method was 0.81 for one of the participants (compared to an accuracy of 0.53 from random predictions).
The number of voxels for which the cross-participant object exemplar identification accuracy was greatest ranged from 50 to 2000 voxels, depending on the participant (Table S1).
The number of voxels (each 3.125×3.125×6 mm3 or 59 mm3 in volume) for which object exemplar identification accuracy was greatest (as plotted in Figure 2) ranged from 25 to 400 voxels, depending on the participant (Table S1).
Figure 8 shows an adaptable workflow for this process, using the identification of the InChI semantic identifier for the word type "chemical" as an exemplar.
The identification of the two graphic symbols which were used as exemplars proceeded as follows.
The information content within a number of individual anatomical regions is sufficient for exemplar and category identification, but the content of the representation appears to be somewhat different across regions.
The identification of exemplars for the continental United States (Cardille and Lambois 2010) used the same basic protocol as that used here; however, the underlying satellite classifications differed substantially between the EOSD (Wulder and others 2008a) and the NLCD (Vogelmann and others 2001).
