Deep Learning Combined with Radiologist’s Intervention Achieves Accurate Segmentation of Hepatocellular Carcinoma in Dual-Phase Magnetic Resonance Images

Table 2

Comparisons of segmentation performance between DFN-R and previous studies.

Studies	Algorithm	Images	Average DSC	Patient number	Journal
Linguraru [8]	SVM	CT	0.74	101	IEEE TMI, 2012
Foruzan [9]	SVM	CT	0.82	35	IJCARS, 2016
Li [12]	CNN	CT	0.80	30	JCC, 2017
Li [13]	CNN	CT	0.74	248	J PERS MED, 2022
Christ [15]	CFCNs	DW-MRI	0.69	31	MICCAI, 2016
Fabijańska [17]	U-Net	DCE-MRI	0.48	9	ICCVG, 2018
Khaled [18]	DCNN	T1-weighted MRI	0.68	174	ABDOM RADIOL, 2020
Zheng [19]	CNN	DCE-MRI	0.83	190	IEEE TMI, 2020
Current study	DFN-R	HBP-MRI and PVP-MRI	0.83	51	—

Notes: MRI: magnetic resonance imaging; DCNN: deep convolutional neural networks; DFN: deep fusion network; SVM: support vector machine; CNN: convolutional neural networks; CFCNs: cascaded fully convolutional neural networks; DW-MRI: diffusion-weighted MRI; DCE-MRI: dynamic contrast-enhanced MRI.