Table 1.
Performance of the software and clinical reports against reference standards, with 95% CI
| Test | Suspicious nodule identified by software | Suspicious nodule identified by software | Any suspicious abnormality on clinical report* |
| Reference standard | Suspicious nodule on clinical report† | Cancer diagnosis by multidisciplinary team decision | |
| Sensitivity | 54.5% (44.2% to 64.4%) | 60.9% (50.1% to 70.9%) | 66.3% (55.7% to 75.8%) |
| Specificity | 83.2% (82.2% to 84.1%) | 83.3% (82.3% to 84.2%) | 98.0% (97.7% to 98.4%) |
| PPV | 5.5% (4.6% to 6.6%) | 5.6% (4.8% to 6.6%) | 35.7% (30.5% to 41.2%) |
| NPV | 99.0% (98.8% to 99.2%) | 99.2% (99.0% to 99.4%) | 99.4% (99.3% to 99.6%) |
| Accuracy | 82.7% (81.7% to 83.6%) | 82.9% (81.9% to 83.9%) | 97.5% (97.1% to 97.9%) |
| FPPI | 0.18 | 0.18 | 0.02 |
*Includes nodules, masses and secondary features of malignancy (such as unilateral pleural effusions or lobar collapse).
†Other suspicious abnormalities (including masses and secondary features of cancer, such as lung volume loss) were excluded.
FPPI, false positives per image; NPV, negative predictive value; PPV, positive predictive value.