Table 1.
Characteristics of the study population and distribution of cancer type, the AI program’s delineation of lesions, lesion risk scores and radiologists’ assessments
| Number | n = 120 | Excluded n = 39 | |
|---|---|---|---|
| Age, years | |||
| Mean ± SD | 62 ± 9 | 63 ± 9 | |
| Median (range) | 65 (44–74) | 66 (44–74) | |
| Cancer typea/ b | IDC | ILC | IDC / ILC |
| Number (%) | 84 (70) | 13 (11) | 29 (74) / 2 (5) |
| Lesion delineationc | Yes | No | |
| Number (%) | 100 (83) | 20 (17) | |
| Overall malignancy risk score | |||
| Mean ± SD | 9,96 ± 0,32 | 7,35 ± 2,48 | |
| Median (range) | 10 (7–10) | 8 (3–10) | |
| Lesion risk score | CC | MLO | |
| Mean ± SD | 74 ± 23 | 66 ± 24 | |
| Median (range) | 85 (27–95) | 71 (25–95) | |
| Radiologists’ assessmentsd | Detected | Missed | Detected / Missed |
| Number (%) | 105 (88) | 15 (13) | 33 (85) / 6 (15) |
| Overall malignancy risk score | |||
| Mean ± SD | 9,68 ± 1,22 | 8,47 ± 2,17 | |
| Median (range) | 10 (3–10) | 10 (3–10) | |
AI, artificial intelligence.
IDC = invasive ductal cancer, ILC = invasive lobular cancer. Other cancer types: invasive tubular cancer (n = 3; 3%), invasive mucinous cancer (n = 1; 1%), invasive papillary cancer (n = 1; 1%), metastasis of non-breast primary cancer (n = 1; 1%) and unknown type (in situ n = 14; invasive n = 3; total n = 17, 14%).
Distribution of cancer types in the excluded group, in addition to IDC and ILC: invasive tubular cancer (n = 1; 3%), papillary cancer (n = 2; 5%) and unknown type (n = 5; 13%).
CC = cranio-caudal projection, MLO = medio-lateral-oblique projection. Missing data: 65 (54%) for CC and 62 (52%) for MLO.
Cancerous lesion detected by both radiologists versusvs missed by one radiologist.