Table 6.

Stepwise backward regression model –icobrain dm for age- and sex-adjusted brain volumes normalized for head size

icobrain dm
Disease Stages	Final Brain Structures	AIC	Δ-AIC	rel.LL	Youden Index	Threshold	Specificity (%)	Sensitivity (%)	AUC
SCD versus HC	WB, FL, PL, and HIP-R	233.7	5.35	0.04	0.213	0.564	81.8	53.4	0.689
McI versus HC	WB, PL, TL, and HIP-R	342.4	4.13	0.13	0.507	0.753	75.0	75.7	0.809
ADD versus HC	WB, PL, HIP-R, HIP-L, and LVENT	215.4	3.05	0.22	0.692	0.620	83.0	86.3	0.914
MCI versus SCD	FL, TL, and HIP-L	397.0	5.24	0.07	0.692	0.829	84.1	48.9	0.691
ADD versus SCD	FL, TL, HIP-R, and HIP-L	276.5	2.98	0.23	0.560	0.673	79.5	76.4	0.831
ADD versus MCI	PL, HIP-R, HIP-L, and LVENT	661.0	3.33	0.19	0.301	0.360	58.0	72.1	0.670

Input brain volumes: whole brain (WB), frontal cortex (FL), parietal cortex (PL), temporal cortex (TL), lateral ventricles (LVENT), left hippocampus (HIP-L), and right hippocampus (HIP-R). Disease stages: HC, cognitively healthy controls; SCD, subjective cognitive decline; MCI, mild cognitive impairment; ADD, Alzheimer’s disease dementia. AIC, Akaike Information Criterion; Δ-AIC, Difference between the AIC of the full model (input brain volumes) and the final model (brain volume model with the lowest AIC); REL.LL, relative likelihood measure; AUC, Area under the curve, specificity (%), and sensitivity (%) values were reported. Specificity and sensitivity outcomes are presented as percentages and were calculated based on the combined smallest Euclidian distance, using the Youden index to determine the threshold. Brain structures highlighted in bold are present in the final models of both automated volumetric tools.