Table 3 Correlation between the significant values and lymph node metastasis.

From: Artificial intelligence-based radiomics for the prediction of nodal metastasis in early-stage lung cancer

Cohort

Solid-part size

pN (%)

p value

Sensitivity (%)

Specificity (%)

Accuracy (%)

NPV (%)

Derivation

> 1.83 cm

44 (23)

< 0.001

44/56 (79)

278/424 (66)

322/480 (67)

278/290 (96)

< 1.83 cm

12 (4)

Validation

> 1.83 cm

16 (15)

0.114

16/27 (59)

121/213 (57)

137/240 (57)

121/132 (92)

< 1.83 cm

11 (8)

All

> 1.83 cm

60 (20)

< 0 .001

60/83 (72)

399/637 (63)

459/720 (64)

399/422 (95)

< 1.83 cm

23 (6)

Cohort

Average solid CT value

pN (%)

p value

Sensitivity (%)

Specificity (%)

Accuracy (%)

NPV (%)

Derivation

>  − 103 HU

39 (22)

< 0.001

39/56 (70)

288/424 (68)

327/480 (68)

288/305 (94)

<  − 103 HU

17 (6)

Validation

>  − 103 HU

18 (17)

0.011

18/27 (67)

126/213 (59)

144/240 (60)

126/135 (93)

<  − 103 HU

9 (7)

All

>  − 103 HU

57 (20)

< 0 .001

57/83 (69)

414/637 (65)

471/720 (65)

414/440 (94)

<  − 103 HU

26 (6)

  1. pN pathological lymph node metastasis; NPV negative predictive value; CT computed tomography; HU Hounsfield units.