Skip to main content

Table 1 Clinical characteristics in the training and test sets for radiomics analysis

From: Hybrid transformer convolutional neural network-based radiomics models for osteoporosis screening in routine CT

 

Training cohort

(n = 204)

 

Test cohort

(n = 79)

 

P

Osteoporosis

Non-osteoporosis

P

Osteoporosis

Non-osteoporosis

P

Sex

 

0.001*

 

0.827

0.499

 Male

3

68

 

6

28

  

 Female

30

103

10

35

Age

63.64.12 ± 7.47

59.13 ± 9.76

0.012*

63.56 ± 9.83

61.03 ± 9.95

0.365

0.063

HU

72.02 ± 33.25

137.10 ± 45.94

P < 0.001*

83.14 ± 41.89

120.10 ± 40.20

0.002*

0.030*

HTCNN_ HU

108.10 ± 29.84

182.90.1 ± 42.24

P < 0.001*

122.90 ± 34.97

167.00 ± 36.48

P < 0.001*

0.039*

  1. Data are mean ± standard deviation.
  2. HTCNN, hybrid transformer deep convolutional neural network; HU, Hounsfield unit.
  3. *P value < 0.05
Back to article page

BMC Medical Imaging

ISSN: 1471-2342

Contact us