Table 4 Review of the papers that applied deep learning for pulmonary vessel segmentation.

References	Modality	Data source	No. of subj.	ML model	Input	DSC	Add’l perf. metrics
Tan et al.⁷³	CT and CTA	Publicly available	16	2D-3D U-Net nnU-net	2D images 3D volume	0.786, 0.797, 0.77 (on CT)	OR = 0.281, 0.285, 0.304
Nam et al.⁷⁴	CTA	Private	104	3D U-Net	3D patches	0.915 ± 0.31	AUC = 0.995
Guo et al.⁷⁵	CT	Private	50	3D CNN	3D patches	0.943	–
Xu et al.⁷⁶	CT	Private		2D CNN		–	–
Nardelli et al.⁷⁷	CT	Publicly available	55	3D CNN	2D and 3D patches	–	Sen = 0.93 Prec = 0.83
Wu et al.⁷⁸	CT	Private	143	MSI-U-Net	3D volume	0.7168	Sen = 0.7234, Prec = 0.7893

CNN convolutional neural network, CT computed tomography, CTA computed tomography angiography, MSI-U-Net multi-scale interactive U-Net, OR over segmentation rate, AUC area under the receiver operating characteristic curve, Sen sensitivity, Spec specificity, Prec precision.

Quick links

Search