Table 3.

Calculation of discriminability index for component identification (DICI).

	Voxels labeled “1” in the component (significant voxels)	Voxels labeled “0” in the component (non-significant voxels)	DICI = z(HR)-z(FAR)a
Voxels labeled “1” in individual tumor template	Hit (true positive)	Miss (false negative)	Hit rate (HR) = hit/voxels labeled “1” in individual tumor template
Voxels labeled “0” in individual tumor template	False alarm (false positive)	Correct rejection (true negative)	False alarm rate (FAR) = false alarm/voxels labeled “0” in individual tumor template

^aThe HR and the FAR was transformed to z-scores according to an inverted cumulative distribution function of a standard Gaussian distribution (mean = 0 and standard deviation = 1).