A deep learning model using chest X-ray for identifying TB and NTM-LD patients: a cross-sectional study

Insights into Imaging

Table 3 Comparison between different deep neural network models on tuberculosis (A) and mycobacteria (B) detection

	Our DNN model			DNN model analogue
	Low TB burden	Medium TB burden	High TB burden	Low TB burden	Medium TB burden	High TB burden
(A)
AUC	0.77 ± 0.14	0.77 ± 0.11	0.77 ± 0.06	0.76 ± 0.11	0.76 ± 0.09	0.77 ± 0.06
Sensitivity	0.62 ± 0.23	0.66 ± 0.20	0.66 ± 0.12	0.64 ± 0.22	0.65 ± 0.16	0.66 ± 0.11
Specificity	0.78 ± 0.04	0.78 ± 0.04	0.77 ± 0.04	0.82 ± 0.03	0.84 ± 0.03	0.84 ± 0.04
PPV	0.09 ± 0.04	0.14 ± 0.04	0.30 ± 0.05	0.11 ± 0.05	0.18 ± 0.05	0.38 ± 0.07
NPV	0.98 ± 0.01	0.98 ± 0.01	0.94 ± 0.02	0.98 ± 0.01	0.98 ± 0.01	0.95 ± 0.02
Reduced further test (%)	77 ± 4	76 ± 4	72 ± 4	81 ± 3	82 ± 3	78 ± 3
Number needs to screen	12.67 ± 5.93	8.33 ± 3.87	3.47 ± 0.57	10.03 ± 4.91	5.98 ± 2.02	2.77 ± 0.62
Misclassified NTM-LD as TB (%)	14 ± 7	16 ± 16	14 ± 15	31 ± 10	36 ± 21	32 ± 21
(B)
AUC	0.77 ± 0.05	0.76 ± 0.07	0.74 ± 0.05	0.67 ± 0.06	0.71 ± 0.07	0.73 ± 0.07
Sensitivity	0.79 ± 0.08	0.78 ± 0.12	0.76 ± 0.09	0.38 ± 0.10	0.49 ± 0.13	0.55 ± 0.11
Specificity	0.61 ± 0.04	0.59 ± 0.05	0.60 ± 0.05	0.85 ± 0.03	0.85 ± 0.04	0.85 ± 0.03
PPV	0.34 ± 0.03	0.18 ± 0.03	0.28 ± 0.03	0.39 ± 0.08	0.27 ± 0.08	0.43 ± 0.08
NPV	0.92 ± 0.03	0.96 ± 0.02	0.93 ± 0.03	0.85 ± 0.02	0.94 ± 0.02	0.91 ± 0.02
Reduced further test (%)	53 ± 4	56 ± 5	54 ± 4	80 ± 3	82 ± 4	79 ± 3
Number needs to screen	2.99 ± 0.30	5.81 ± 1.11	3.67 ± 0.47	2.70 ± 0.60	4.07 ± 1.45	2.40 ± 0.49
Misclassified TB as imitator (%)	26 ± 21	26 ± 17	26 ± 11	38 ± 24	33 ± 19	37 ± 14
Misclassified NTM-LD as imitator (%)	20 ± 8	18 ± 15	21 ± 17	67 ± 10	70 ± 19	68 ± 19

DNN, deep neural network; TB, tuberculosis; AUC, area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value; NTM-LD, nontuberculous mycobacterial lung disease