Skip to main content

Table 2 NLST test set nodule characteristics

From: High precision localization of pulmonary nodules on chest CT utilizing axial slice number labels

Nodule variable

Missed/incorrect/unintended nodule predictions

n (%)

Correct nodule predictions n

(%)

p

Totals

n (%)

N

35 (100%)

54 (100%)

–

89 (100%)

Location

  

0.307

 

 Left lower lobe

7 (20%)

10 (19%)

 

17 (19%)

 Left upper lobe

6 (17%)

7 (13%)

 

13 (15%)

 Lingula

0 (0%)

6 (11%)

 

6 (7%)

 Right lower lobe

12 (34%)

12 (22%)

 

24 (27%)

 Right middle lobe

4 (11%)

10 (19%)

 

14 (16%)

 Right upper lobe

6 (17%)

9 (17%)

 

15 (17%)

Central versus peripheral

  

0.332

 

 Central

3 (9%)

1 (2%)

 

4 (4%)

 Peripheral

32 (91%)

53 (98%)

 

85 (96%)

Subpleural versus parenchymal

  

0.807

 

 Subpleural

19 (54%)

32 (59%)

 

38 (43%)

 Parenchymal

16 (46%)

22 (41%)

 

51 (57%)

Margins

  

0.254

 

 Smooth

27 (77%)

46 (85%)

 

73 (82%)

 Poorly defined

5 (14%)

5 (9%)

 

10 (11%)

 Spiculated

1 (3%)

3 (6%)

 

4 (4%)

 Unable to determine

2 (6%)

0 (0%)

 

3 (2%)

Diameter (mm)

  

0.752

 

 (mean ± SD)

4.77 ± 2.78

4.94 ± 1.75

 

4.88 ± 2.26

Attenuation

  

0.028*

 

 Soft tissue

27 (77%)

51 (94%)

 

78 (88%)

 Ground glass

3 (9%)

1 (2%)

 

4 (4%)

 Mixed

1 (3%)

2 (4%)

 

4 (4%)

 Unable to determine

4 (11%)

0 (0%)

 

3 (3%)

  1. Adjusted nodule performance with the highest recall score at a 20 mm distance threshold was used to split missed/incorrect/unintended and correct nodule predictions
  2. SD standard deviation
  3. *p < 0.05 using Chi-square test for categorical and Welch’s T-test for continuous variables to test for difference between correct and missed/incorrect nodule predictions