26 _________________________________________________________________________________Bota et al.

answers. Rules for inferring qualitative spatial

tABi if t ABi = tABl∀i , l = 1,…n ; l ≠ i
m relations when the completeness constraint is
if tABi = m ∧ tABl = d
 ∀i , l = 1,… n; l ≠ i not fulfilled are given in Eq. (2)
cvBy if t = i ∧ t
 ABi ABl
gtAB = = {cvBy,isCo} {i,cv,cvBy,o} if tABi = i ∀i = 1,… , n
 t ,o if tABi = {cv,cvBy}
{ AB }
 = ∀l = 1,…n ; l ≠ i
cv if tABi = i ∧ t ABl = {cv,co} ∀i = 1,… , n
 (1) 
∀l = 1,…n ; l ≠ i { }
 cv,co,o if t ABi = co ∀i = 1,… , n
0 otherwise {cvBy,isCo,o} if tABi = isCo
 ∀i = 1,… , n

{ } ABi = {m,d}
where “^” stands for logical AND. d, m,o if t
The general topological relation between  ∀i = 1,… , n
regions A and B (gtAB) is of the same type with g t AB = {m, o} if tABi = d ∧ ∃ tABl = m
any of the topological relations {tABi} estab-  ∀i = 1,… , n, l ≠ i
lished across a set of Atlas Level pairs if all ele- {cvBy,o} if ∃ tABi = i ∧ tABl

ments of the set are identical. If the set of  = {cvBy,isCo}
topological relations contains at least one  ∀l = 1,… , n; l ≠ i
“meet” relation and the rest are all disjoint, then {cv,o} if ∃ tABi = i ∧ tABl
the inferred general spatial relation is “meet.”  = {cvBy,isCo}

The converse of this rule is also true: the gen-  ∀l = 1,… , n; l ≠ i
eral spatial relation is “meet” if there is at least 0 otherwise
one “disjoint” topological relation and the
(2)
remaining relations are “meet.” The general
topological relation is considered to be “cov- The results of the first rule in Eq. (1) become
ered by” when the set of topological relations equivocal in Eq. (2) because the set of compared
is formed by the relations “identical” and either Atlas Levels is not complete. Therefore, rela-
“covered by” or “is contained.” The converse tionships between the interiors and boundaries
of this rule is also true. The general topologi- of the brain regions established across Atlas
cal relation is considered to be “overlap” if there Levels are not necessarily true in those pairs of
is at least one topological relation “overlap” in Atlas Levels that are missing from the com-
the set {tABi}, or the set {tABi} contains at least plete set. For example, the result of the first rule
one “covers” or “contains” relation and at least in Eq. (2) can be “identical,” “covers,” “cov-
one complementary relation (cvBy, or isCo, ered by,” or “overlap,” even though all topo-
respectively). The general relation “overlap” is logical relations are “identical.” Because the set
also inferred when there is one “disjoint,” or of Atlas Levels is incomplete, one must assume
“meet” relation and the rest are any combina- that there may be a pair of Atlas Levels where
tion of the set {i, co, cv, isCo, cvBy}. the relationship between the compared regions
Brain regions defined in different neu- is “disjoint” or “meet.” Thus, the general topo-
roanatomical nomenclatures may be difficult logical relation is not only “identical,” but it
to compare topologically because of their com- can also be “overlap,” because the boundaries
plex spatial relations, or because there is a of the compared region intersect in at least one
paucity of landmarks, and thus the complete- pair of Atlas Levels. In a similar way, the result
ness constraint defined earlier may not be sat- of the third rule can be “covers,” “contains,”
isfied. In this case, the inference of topological or “overlap” when all the topological relations
relations does not always yield unequivocal are “covers.” One may assume that there is a

Neuroinformatics _______________________________________________________________ Volume 3, 2005