Table 11.

Summary of two-stage strategies in deep learning-based methods.

Study	Year	Datasets	No. of Patients	Performance			Image Modalities	Time/ Equipment	Category
				Dice (%)	HD (mm)	95HD (mm)
Zhang [97]	2021	In-house	170	$89.00\pm 2.00$	—	$1.66\pm 0.51$	CT	40.1 s/GPU	DL
Tappeiner [98]	2019	PDDCA	40	$91.00\pm 2.00$	—	$2.4\pm 0.6$	CT	38.3 s/GPU	DL
Mu [99]	2020	In-house	50	$89.80\pm 2.70\left(L\right)$; $90.40\pm 2.00\left(R\right)$	—	—	CT	3 s/GPU	DL
Wang [100]	2018	PDDCA	48	$93.00\pm 1.90$	—	$1.26\pm 0.50$	CT	6 s/GPU	DL
Tang [32]	2019	In-house HNC+ HNPETCT PDDCA	175 35 + 105 48	$93.12\pm 1.41$ $89.31\pm 11.59$ $95.00\pm 0.80$	—	$2.48\pm 0.83$ $3.05\pm 2.60$ —	CT	2 s/GPU	DL
Lei [101]	2020	In-house	15	$85.00\pm 4.00$	—	—	MRI	—	DL
Lei [102]	2020	In-house	15	$88.00\pm 3.00$	—	—	CT	—	DL
Liang [103]	2019	In-house	185	$91.40\pm 0.04\left(L\right)$; $91.20\pm 3.00\left(R\right)$	—	—	CT	30 s/GPU	DL
Dijk [104]	2020	In-house	693	$94.00\pm 1.00$	—	$1.30\pm 0.50$	CT	—	DL
Men [105]	2019	HNSCC	100	$92.00\pm 2.00$	$2.40\pm 0.40$	—	CT	5.5 min/GPU	DL
Egger [106]	2018	In-house	20	$89.64\pm 1.69$	—	—	CT	—/CPU	DL