Research Article

Deep Learning to Predict EGFR Mutation and PD-L1 Expression Status in Non-Small-Cell Lung Cancer on Computed Tomography Images

Table 1

Demographic and clinical characteristics of included NSCLC patients.

CharacteristicsTotal (N = 1262)Training set (N = 882)Validation set (N = 125)Test set (N = 255)p value

Age at diagnosis (year)
 Mean ± SD57.70 ± 10.5057.69 ± 10.2757.48 ± 9.7057.85 ± 10.270.95
Sex
 Male620 (49.13)438 (49.66)66 (52.80)116 (45.49)0.35
 Female642 (50.87)444 (50.34)59 (47.20)139 (54.51)
Smoking
 Current/former452 (35.82)323 (36.62)45 (36.00)84 (32.94)0.18
 Never749 (59.35)520 (58.96)77 (61.60)152 (59.61)
 Unknown61 (4.83)39 (4.42)3 (2.40)19 (7.45)
Gene mutation status
 EGFR(−) PD-L1(−)276 (21.87)193 (21.88)28 (22.40)55 (21.57)1.00
 EGFR(−) PD-L1(+)290 (22.98)203 (23.02)28 (22.40)59 (23.14)
 EGFR(+) PD-L1(−)502 (39.78)350 (39.68)50 (40.00)102 (40.00)
 EGFR(+) PD-L1(+)194 (15.37)136 (15.42)19 (15.20)39 (15.29)
EGFR-TKI-targeted therapy
 Yes391 (30.98)265 (30.05)45 (36.00)81 (31.76)0.39
 No871 (69.02)617 (69.95)80 (64.00)174 (68.24)
ICI therapy
 Yes15 (1.19)12 (1.36)0 (0)3 (1.18)0.42
 No1247 (98.81)870 (98.64)125 (100.00)252 (98.82)
Histopathology
 LUAD1185 (93.90)824 (93.42)119 (95.20)242 (94.90)0.82
 LUSC53 (4.20)41 (4.65)4 (3.20)8 (3.14)
 Others24 (1.90)17 (1.93)2 (1.60)5 (1.96)
Tumor stage
 I529 (41.91)378 (42.86)43 (34.40)108 (42.35)0.22
 II96 (7.61)60 (6.80)9 (7.20)27 (10.59)
 III236 (18.70)160 (18.14)30 (24.00)46 (18.04)
 IV367 (29.08)262 (29.71)37 (29.60)68 (26.67)
 Unknown34 (2.69)22 (2.49)6 (4.80)6 (2.35)
Follow-up
 Median follow-up time (month, 95% CI)31 (30–31)30 (30–31)31 (29–35)32 (30,33)0.09
Overall survival
 Death412 (32.65)283 (32.08)51 (40.80)78 (30.59)0.20
 Median OS (month, 95% CI)44 (42–49)43 (41–49)41 (33-NA)46 (40-NA)

LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma; SD: standard deviation; OS: overall survival; CI: confidence interval.