Skip to main content

Table 1 Clinical descriptors of the derivation and validation cohorts at the baseline

From: Artificial neural networks improve LDCT lung cancer screening: a comparative validation study

  Derivation cohort (N = 602) Validation cohort (N = 234) P b
Cancer (N = 19) Control c (N = 583) Cancer (N = 8) Control c (N = 226)  
Sex a
 Male 7 (36.84%) 236 (40.48%) 2 (25.00%) 111 (49.12%) 0.038
 Female 12 (63.16%) 347 (59.52%) 6 (75.00%) 115 (50.88%)  
Age (y) a 64.89 ± 7.53 61.87 ± 6.42 57.63 ± 8.73 61.05 ± 7.88 0.053
LDCT parameters
 Dose (mSv) a 1.95 ± 0.64 1.46 ± 0.24 1.78 ± 0.45 1.49 ± 0.26 0.161
 DLP (mGy.cm) a 75.53 ± 32.54 49.17 ± 11.08 64.50 ± 24.43 50.77 ± 10.72 0.206
Pattern of nodules
 Nodules of interest a 2.42 (1–7) 1.11 (0–32) 1.88 (1–7) 1.29 (0–8) 0.330
 Number of involved lobes a 1.68 (1–3) 0.75 (0–5) 1.38 (1–4) 0.99 (0–5) 0.007
Size of nodules (mm)
 Solid nodule a 10.01 (0–136.00) 1.49 (0–19.80) 0.63 (0–5.00) 1.80 (0–36.75) 1.000
 PS nodule a 3.89 (0–20.40) 0.38 (0–11.95) 1.77 (0–4.90) 0.54 (0–7.30) 0.498
 GGN a 8.87 (0–31.00) 0.58 (0–23.30) 4.56 (0–10.30) 0.24 (0–9.05) 0.038
 Calcified nodule a 0.00 (0) 0.39 (0–19.25) 0.86 (0–6.90) 0.55 (0–7.05) 0.100
 Fat-containing nodule a 0.00 (0) 0.05 (0–28.15) 0.00 (0) 0.00 (0) 0.506
Intra-pulmonary findings
 Linear atelectasis a 10 (52.63%) 431 (73.93%) 2 (25.00%) 108 (47.79%) < 0.001
 Plate-like atelectasis a 5 (26.32%) 73 (12.52%) 0 (0.00%) 19 (8.41%) 0.050
 Plate-like GGN a 2 (10.53%) 143 (24.53%) 1 (12.50%) 39 (17.26%) 0.029
 Bronchiectasis a 0 (0.00%) 39 (6.69%) 1 (12.50%) 8 (3.54%) 0.143
 Emphysema a 1 (5.26%) 51 (8.75%) 2 (25.00%) 28 (12.39%) 0.068
 Fibrotic change a 2 (10.53%) 154 (26.42%) 0 (0.00%) 42 (18.58%) 0.015
Extra-pulmonary findings
 Mediastinal tumour a 4 (21.05%) 30 (5.15%) 1 (12.50%) 8 (3.54%) 0.290
 Thyroid nodule a 1 (5.26%) 19 (3.26%) 0 (0.00%) 2 (0.88%) 0.045
 Adrenal nodule a 1 (5.26%) 5 (0.86%) 0 (0.00%) 0 (0.00%) 0.125
 Hepatic nodule a 1 (5.26%) 67 (11.49%) 0 (0.00%) 20 (8.85%) 0.245
 Renal nodule a 0 (0.00%) 16 (2.74%) 0 (0.00%) 10 (4.42%) 0.229
Lung-RADS
 Category 1 0 (0.00%) 323 (55.40%) 0 (0.00%) 115 (50.89%) 0.240
 Category 2 6 (31.58%) 222 (38.08%) 7 (87.50%) 102 (45.13%) 0.021
 Category 3 5 (26.32%) 31 (5.32%) 1 (12.50%) 6 (2.65%) 0.080
 Category 4 8 (42.10%) 7 (1.20%) 0 (0.00%) 3 (1.33%) 0.279
  1. a The 22 input features for developing the ANN
  2. b Comparison of the derivation cohort and validation cohort, P-values less than 0.05 indicated statistical significance
  3. c Participant who did not have confirmed lung cancer prior to the index date were labelled as control
  4. BMI body mass index; DLP dose length product; GGN ground-glass nodule; PS nodule, part-solid nodule
  5. The values are given as the mean ± SD, range or n (%)