A dual deep neural network for auto-delineation in cervical cancer radiotherapy with clinical validation

Radiation Oncology

Table 1 A six-point evaluation criterion for the clinical target volume (CTV) delineation in cervical cancer radiotherapy

Nos.	Criterion	Qualitative evaluation (60 cases)
Nos.	Criterion	UNet	SegNet	SegNet^(A)	SegNet^(B)	SegNet^(C)
1	The common iliac lymph nodes area is included	56	60	60	59	60
2	The internal iliac lymph nodes area is included	59	60	60	60	60
3	The external iliac lymph nodes area is included	52	59	57	58	55
4	The presacral lymph nodes area is included	40	60	52	50	43
5	The paravaginal tissue area is included	33	53	45	46	42
6	The upper vagina is included	60	60	60	60	60

Main differences between UNet and SegNet rely on different network architectures. The differences between SegNet, SegNet^(A), SegNet^(B) and SegNet^(C) are the scale of the training set, which is 121 (multisource dataset), 42, 40 and 39, respectively

ISSN: 1748-717X