TABLE 5.

ML applications for brain-age prediction in epilepsy.

References	Subjects	Imaging modality	Imaging features	Classifiers	Main outcomes
Pardoe et al. (2017)	136 FE (94 DR, 42 ND), 74 HC, (2001 HC for model)	T1	VBM	GPR	+4.5 years in DR-FE, but non-significance in ND-FE.
Chen et al. (2019)	35 TLE (17 R, 18 L), 37 HC (300 HC for model)	DSI	GFA, AD, RD, MD, NG, NGO, NGP	GPR	+10.9 years in RTLE, +2.2 years in LTLE Correlation with onset age, duration, seizure frequency
Hwang et al. (2020)	104 TLE, 151 HC	T1, rs-fMRI	SBM, FC	SVR	+6.6 years in structural MRI, +8.3 years in functional MRI
Sone et al. (2021)	318 EPI, 1192 HC	T1	VBM	SVR	>+4 years in almost all forms of epilepsies +10.9 years in TLE with psychosis

AD, axial diffusivity; DR, drug-resistant; DSI, diffusion spectrum imaging; EPI, epilepsy; FC, functional connectivity; FE, focal epilepsy; GFA, generalized fractional anisotropy; GPR, Gaussian process regression; HC, healthy controls; L, left; MD, mean diffusivity; ND, newly diagnosed; NG, non-Gaussianity; NGO, NG orthogonal; NGP, NG parallel; R, right; RD, radial diffusivity; rs-fMRI, resting-state functional MRI; SBM, surface-based morphometry; SVR, support vector regression; TLE, temporal lobe epilepsy; VBM, voxel-based morphometry.