Table 2.

Comparisons of segmentation performance between our proposed CNN model and the similar studies.

Studies	Algorithm	Images used	Average DSCa	Patient number	Journal
Deng et al. (10)	SVMb	DCE-MRIc	0.862	120	Contrast Media and Molecular Imaging, 2018
Song et al. (8)	Graph-based cosegmentation	PET	0.761	2	IEEE Transactions on Medical Imaging, 2013
Yang et al. (9)	MRFsd	PET, CT, MRI	0.740	22	Medical Physics, 2015
Stefano et al. (11)	AK-RWe	PET	0.848	18	Medical and Biological Engineering and Computing, 2017
Wang et al. (4)	CNNf	MRI	0.725	15	Neural Processing Letters, 2018
Ma et al. (12)	CNNs+3D graph cut	MRI	0.851	30	Experimental and Therapeutic Medicine, 2018
Men et al. (15)	DDNNg	CT	0.716	230	Frontiers in Oncology, 2017
Li et al. (16)	CNN	CE-MRI	0.890	29	Biomed Research International, 2018
Huang et al. (17)	CNN	PET-CT	0.736	22	Contrast Media and Molecular Imaging, 2018
Ma et al. (18)	C-CNNh	CT-MRI	0.746	90	Physics in Medicine and Biology, 2019
Proposed method	CNN	Dual-sequence MRI	0.721	44	–

^aDSC, Dice similarity coefficient;

^bSVM, support vector machine;

^cDCE-MRI, dynamic contrast-enhanced magnetic resonance imaging;

^dMRFs, Markov random fields;

^eAK-RW, adaptive random walker with k-means;

^fCNN, convolutional neural network;

^gDDNN, deep deconvolutional neural network;

^hC-CNN, combined convolutional neural network.