Brain Age Estimation on a Dementia Cohort Using FLAIR MRI Biomarkers

Article Figures & Data

Download figure
Open in new tab
Download powerpoint
FIG 1.
Corrected predicted brain age versus true age for NC, sMCI, cMCI, and AD using the RFR.
Download figure
Open in new tab
Download powerpoint
FIG 2.
BrainAGE distributions.
Download figure
Open in new tab
Download powerpoint
FIG 3.
BrainAGE comparison across time after baseline scans for all cognitive groups.

Table 1:

FLAIR MR imaging data sets^{^a}

Table 2:

Brain age prediction model comparison based on MAE and correlation between true and predicted age^{^a}

Model	MAE Raw	MAE Corr.	Correlation Raw	Correlation Corr.
RFC	3.16	2.89	0.70 (P < .001)	0.84 (P < .001)
SVM (reg.)	3.54	2.66	0.66 (P < .001)	0.86 (P < .001)
Log. Reg.	4.06	3.10	0.56 (P < .001)	0.79 (P < .001)
RFR	3.21	2.46	0.74 (P< .001)	0.87 (P < .001)

Note:—corr. indicates corrected; reg., regression; Log. Reg, Logistic Regression.
↵a Results using the raw and corr. brain ages are shown for comparison.

Table 4:

ANOVA comparing slope and intercept of BrainAGE versus time^{^a}

Comparison	Slope Analysis (P Value) (Cohen’s d)	Intercept Analysis (P Value) (Cohen’s d)
NC vs sMCI	.66 (0.43)	.77 (0.30)
NC vs cMCI	.02 (2.29)	.29 (0.26)
NC vs AD	.16 (1.42)	<.01 (8.65)
sMCI vs cMCI	.04 (3.00)	.50 (0.18)
sMCI vs AD	.06 (1.90)	<.01 (9.03)
cMCI vs AD	.69 (0.42)	<.01 (7.02)

Table 6:

Results summary from the brain age prediction model

Experiment	Results	Notes
Model validation	MAE = 2.46 yr	Existing work MAE = 4–13.5 yr
Cross-sectional ANOVA	P < .01	All comparisons except NC/sMCI
Longitudinal ANOVA (slopes)	P < .05	All cMCI comparisons except cMCI-AD
Longitudinal ANOVA (intercepts)	P < .01	All AD comparisons
1-year interval t tests	P < .01	sMCI/cMCI up to 4 years before conversion
Correlation analysis	P < .05	p-tau, aβ, APOE-4