notesum.ai

Published at October 22

IdenBAT: Disentangled Representation Learning for Identity-Preserved Brain Age Transformation

cs.LG

cs.AI

Released Date: October 22, 2024

Authors: Junyeong Maeng, Kwanseok Oh, Wonsik Jung¹, Heung-Il Suk²

Aff.: ¹Department of Artificial Intelligence, Korea University, Seoul, 02841, Republic of Korea; ²Department of Brain and Cognitive Engineering, Korea University, Seoul, 02841, Republic of Korea

Arxiv: https://arxiv.org/abs/2410.16945v1

Dimension	Methods	Brain Aging Evaluation			Age Transformation Accuracy (PAD ( $\downarrow$ ))
(2D / 3D)	Methods	PSNR ( $\uparrow$ )	SSIM ( $\uparrow$ )	MSE ( $\downarrow$ )	All	(48 - 54)	(55 - 61)	(62 - 68)	(69 - 74)	(75 - 80)
	cGAN Mirza and Osindero (2014)	23.835	0.905	0.0043	7.78	5.21	3.41	5.07	7.24	12.72
	CAAE Zhang et al. (2017)	18.743	0.660	0.0135	7.53	3.39	3.18	5.69	7.92	11.11
2D	LSAB Xia et al. (2021)	26.186	0.604	0.0092	5.39	4.30	4.00	5.02	5.05	6.73
	CBAS Xia et al. (2019)	26.171	0.790	0.0096	5.84	3.49	3.89	4.99	5.62	7.90
	IdenBAT (Ours)	26.314	0.924	0.0026	3.52	3.27	3.05	3.56	3.14	4.72
3D	CounterSynth Pombo et al. (2023)	22.879	0.771	0.0062	9.06	13.15	7.45	4.77	7.28	12.96
3D	IdenBAT (Ours)	29.653	0.958	0.0020	4.68	5.24	3.47	4.26	5.27	5.35