Fully Automated Delineation of Gross Tumor Volume for Head and Neck Cancer on PET-CT Using Deep Learning: A Dual-Center Study

Table 1

Comparison of the segmentation performance in Experiment 2 between our proposed CNN model and the similar studies.

Studies	Algorithm	Images used	Average DSC	Automatic	Patient number	Center number	Journal
Yang et al. [13]	MRFs	PET, CT, and MRI	0.740	Fully automatic	22	1	Medical Physics
Stefano et al. [14]	AK-RW	PET	0.848	Semiautomatic	18	1	Medical & Biological Engineering & Computing
Berthon et al. [15]	Decision tree	PET	0.770	Fully automatic	20	1	Radiotherapy & Oncology
Song et al. [17]	Graph-based cosegmentation	PET	0.761	Semiautomatic	2	1	IEEE Transactions on Medical Imaging
Zeng et al. [16]	Active surface modeling	PET	0.700	Fully automatic	2	1	Computers in Biology and Medicine
Proposed method	CNN	PET and CT	0.736	Fully automatic	22	2	—

Note: DSC, dice similarity coefficient; MRFs, Markov random fields; AK-RW, adaptive random walker with k-means; CNN, convolutional neural network.