Fully automated pelvic bone segmentation in multiparameteric MRI using a 3D convolutional neural network

Insights into Imaging

Table 2 Demographics of patients among different dataset

Characteristic	Algorithm development^# (N = 264)			External Set* (N = 60)	F Value	P value
Characteristic	Training set	Validation set	Testing set	External Set* (N = 60)	F Value	P value
No. of patients	208	28	28	60	–	–
Age (mean, years) (SD)	67.13 (9.75)	67.03 (9.93)	67.21 (9.35)	65.38 (13.07)	0.458	0.712
PSA (median, ng/ml)
T-PSA (range)	10.34 (0.15, 156.00)	9.01 (0.77, 128.3)	8.96 (0.82, 27.99)	8.14 (1.38, 50.00)	1.349	0.259
F-PSA (range)	1.32 (0.09, 23.46)	1.08 (0.46, 9.05)	1.32 (0.34, 3.53)	1.43 (0.33, 13.20)	0.342	0.795
F/T-PSA (range)	0.13 (0.02, 0.36)	0.13 (0.05, 0.26)	0.16 (0.05, 0.24)	0.14 (0.05, 0.65)	2.077	0.104

SD standard deviation, PSA prostate specific antigen, T-PSA total PSA, F-PSA free PSA
^#The data in algorithm development were collected between August 2018 and August 2019
^*The data in external validation were collected between January 2020 and March 2020