Skip to main content

Table 4 Detection rates across cancer subgroups for screen-detected cancers and interval cancers in the Standalone AI scenario

From: Breast cancer detection accuracy of AI in an entire screening population: a retrospective, multicentre study

 

First reader

Standalone AIsens

Standalone AIspec

Screen-detected cancers (n = 1268)*

 

Tumour size

   

0–10 mm (n = 487)

11–20 mm (n = 574)

21–50 mm (n = 179)

51 + mm (n = 14)

Unknown (n = 14)

390 (80.1); ref.

510 (88.9); ref.

161 (89.9); ref.

11 (78.6); ref.

14 (100.0); ref.

356 (73.1); 0.007

465 (81.0); 0.0001

142 (79.3); 0.007

13 (92.9); 0.33

7 (50.0); <0.0001

325 (66.7); <0.0001

436 (76.0); <0.0001

137 (76.5); 0.001

13 (92.9); 0.33

6 (42.9); <0.0001

Malignancy grade

   

Grade 1 (n = 403)

Grade 2 (n = 576)

Grade 3 (n = 192)

Unknown (n = 97)

323 (80.2); ref.

505 (87.7); ref.

180 (93.8); ref.

78 (80.4); ref.

335 (83.1); 0.27

449 (78.0); <0.0001

142 (74.0); <0.0001

57 (58.8); 0.001

303 (75.2); 0.06

427 (74.1); <0.0001

133 (69.3); <0.0001

54 (55.7); 0.0002

TNM stage

   

Local (I + II) (n = 1253)

Locally advanced (III) (n = 11)

Distant metastasis (IV) (n = 4)

1071 (85.5); ref.

11 (100.0); ref.

4 (100.0); ref.

971 (77.5); <0.0001

8 (72.7); <0.0001

4 (100.0); >0.99

906 (72.3); <0.0001

8 (72.7); <0.0001

3 (75.0); <0.0001

Lymph node positivity

   

No (n = 964)

Yes (n = 304)

814 (84.4); ref.

272 (89.5); ref.

743 (77.1); <0.0001

240 (79.0); 0.0004

692 (71.8); <0.0001

225 (74.0); <0.0001

ER positivity

   

0% (n = 94)

1–9% (n = 47)

10–100% (n = 1120)

Unknown (n = 7)

88 (93.6); ref.

44 (93.6); ref.

948 (84.6); ref.

6 (85.7); ref.

60 (63.8); <0.0001

28 (59.6); 0.001

890 (79.5); 0.001

5 (71.4); >0.99

56 (59.6); <0.0001

24 (51.1); 0.0001

832 (74.3); <0.0001

5 (71.4); >0.99

HER2 status

   

Negative (n = 1127)

Positive (n = 130)

Unknown (n = 11)

959 (85.1); ref.

118 (90.8); ref.

9 (81.8); ref.

869 (77.1); <0.0001

108 (83.1); 0.09

6 (54.6); 0.38

809 (71.8); <0.0001

102 (78.5); 0.005

6 (54.6); 0.38

Interval cancers (n = 562)*

 

Tumour size

   

0–10 mm (n = 90)

11–20 mm (n = 216)

21–50 mm (n = 201)

51 + mm (n = 35)

Unknown (n = 20)

9 (10.0); ref.

11 (5.1); ref.

13 (6.5); ref.

6 (17.1); ref.

0 (0.0); ref.

23 (25.6); 0.004

48 (22.2); <0.0001

52 (25.9); <0.0001

13 (37.1); 0.04

3 (15.0); <0.0001

17 (18.9); 0.12

39 (18.1); <0.0001

40 (19.9); <0.0001

10 (28.6); 0.22

3 (15.0); <0.0001

Malignancy grade

   

Grade 1 (n = 104)

Grade 2 (n = 239)

Grade 3 (n = 166)

Unknown (n = 53)

8 (7.7); ref.

15 (6.3); ref.

13 (7.8); ref.

3 (5.7); ref.

24 (23.1); 0.003

77 (32.2); <0.0001

32 (19.3); 0.001

6 (11.3); 0.45

21 (20.2); 0.01

60 (25.1); <0.0001

24 (14.5); 0.04

4 (7.6); >0.99

TNM stage

   

Local (I + II) (n = 508)

Locally advanced (III) (n = 33)

Distant metastasis (IV) (n = 16)

Unknown (n = 5)

34 (6.7); ref.

4 (12.1); ref.

0 (0.0); ref.

1 (20.0); ref.

129 (25.4); <0.0001

6 (18.2); 0.69

3 (18.8); <0.0001

1 (20.0); >0.99

100 (19.7); <0.0001

5 (15.2); >0.99

3 (18.8); <0.0001

1 (20.0); >0.99

Lymph node positivity

   

No (n = 376)

Yes (n = 186)

26 (6.9); ref.

13 (7.0); ref.

83 (22.1); <0.0001

56 (30.1); <0.0001

67 (17.8); <0.0001

42 (22.6); <0.0001

ER positivity

   

0% (n = 113)

1–9% (n = 51)

10–100% (n = 394)

Unknown (n = 4)

8 (7.1); ref.

2 (3.9); ref.

29 (7.4); ref.

0 (0.0); ref.

15 (13.3); 0.07

10 (19.6); 0.02

113 (28.7); <0.0001

1 (25.0); <0.0001

11 (9.7); 0.58

9 (17.7); 0.04

88 (22.3); <0.0001

1 (25.0) < 0.0001

HER2 status

   

Negative (n = 454)

Positive (n = 95)

Unknown (n = 13)

33 (7.3); ref.

5 (5.3); ref.

1 (7.7); ref.

117 (25.8); <0.0001

20 (21.1); 0.0003

2 (15.4); 0.26

93 (20.5); <0.0001

14 (14.7); 0.04

2 (15.4); 0.26

  1. Data are n (%); p value. The cancer detection rate is reported as the number of detected cancers out of the number of true cancers for the subgroup in the same row. TNM = tumour, node, metastasis. ER = estrogen receptor. HER2 = human epidermal growth factor receptor 2. AIsens=artificial intelligence score cut-off point matched at mean first reader sensitivity. AIspec=artificial intelligence score cut-off point matched at mean first reader specificity. *Reported for invasive cancers only. †Exact binomial test used instead of McNemar’s test due to small discordant cells