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

BMC Medical Imaging

Table 3 The discriminative power of osteoporosis for four predictors

Predictors	Group	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
HU	Train	0.88 (0.82–0.93)	0.76	0.73	0.91	0.98	0.40
HU	Test	0.74 (0.59–0.89)	0.85	0.94	0.50	0.89	0.67
HTCNN_ HU	Train	0.94 (0.91–0.98)	0.86	0.84	0.94	0.99	0.53
HTCNN_ HU	Test	0.79 (0.67–0.91)	0.58	0.48	1.00	1.00	0.33
Trabecular rad-score	Train	0.93 (0.88–0.98)	0.87	0.88	0.87	0.57	0.97
Trabecular rad-score	Test	0.84 (0.74–0.94)	0.73	0.88	0.70	0.42	0.96
Vertebrae rad-score	Train	0.95 (0.91–0.99)	0.94	0.85	0.96	0.80	0.97
Vertebrae rad-score	Test	0.97 (0.93-1.00)	0.91	0.94	0.91	0.71	0.98

PPV, positive predictive value; NPV, negative predictive value; HTCNN, hybrid transformer deep convolutional neural network; HU, Hounsfield unit

ISSN: 1471-2342