Development and evaluation of a deep learning framework for pelvic and sacral tumor segmentation from multi-sequence MRI: a retrospective study

Table 1 Clinical characteristic of patients

	Train^a(n = 400)	Validation^b(n = 100)	Test^c(n = 116)	p-value
Age	41.83 ± 18.11	39.89 ± 16.78	43.04 ± 16.68	0.436
Size	10·00 ± 4.44	9.95 ± 4.61	9.93 ± 3.83	0.765
Sex				0.414
Male	219	56	56
Female	181	44	60
Location				0.544
Iliac crest	80	22	26
Acetabulum	27	12	5
Pubis/ischium	32	6	7
Sacrum	206	45	63
Multi-zone^d	55	15	15
Malignancy				0.799
Malignant	302	78	86
Benign	98	22	30

Note:
a, In the training set, the other tumor locations include 80 on iliac crest, 55 on multi-zone, 32 on pubis/ischium and 27 on acetabulum
b, In the validation set, the other tumor locations include 22 on iliac crest, 15 on multi-zone, 12 on acetabulum and 6 on pubis/ischium
c, In the test set, the other tumor locations include 26 on iliac crest, 15 on multi-zone, 7 on pubis/ischium and 5 on acetabulum
d, “multi-zone” is a tumor that involves more than one area simultaneously

ISSN: 1470-7330