ID-Seg: an infant deep learning-based segmentation framework to improve limbic structure estimates

Brain Informatics

Table 1 Demographics and MRI sequence information

	Training DHCP^* (N = 473)	ECHO-Dataset1^ƒ (N = 20)	M-CRIB^† (N = 10)	ECHO-Dataset2^‡ (N = 50)
PMA at scan, weeks	40.65 ± 2.19	46.90 ± 4.14	39.78 ± 1.31	48.06 ± 4.69
Sex
Female, N(%)	266 (43.8%)	13 (65.0%)	4 (40.0%)	25 (50%)
Male, N (%)	207 (56.2%)	7 (35.0%)	6 (60.0%)	25 (50%)
MRI scanners	3 Philips	3T GE	3T Siemens	3T GE
MRI resolution (mm³)	[0.5, 0.5, 0.5]	[0.9, 0.9, 0.9]	[0.63, 0.63, 0.63]	[0.9, 0.9, 0.9]
MRI dimensions	[290,290,203]	[130,256,256]	[304,304,157]	[130, 256, 256]

For quantitative variables, data are presented as mean ± standard deviation unless otherwise noted. PMA: postmenstrual age. ^*The large training DHCP dataset with corresponding dHCP labels was used to pre-train the model with sufficient data. ^ƒThe ECHO-Dataset1 dataset was used to test the model’s performance as an internal source. ^†The M-CRIB dataset was used as an external test dataset to further test the reliability of our proposed deep learning framework. ^‡The proof-of-concept ECHO-Dataset2 was used to test the association between brain morphometric measures at birth and corresponding CBCL measures at age 2-year-old.