Table 1.
A summary of the articles in the literature review that applied deep learning techniques for pulmonary embolism detection on computed tomographic pulmonary angiography.
| Author | Year | Study design | Database type | Dataset size (n = studies) | Images evaluated by | Performance scores |
|---|---|---|---|---|---|---|
| Huang et al.27 | 2020 | Retrospective | Proprietary | 1997 | Board-certified radiologist |
AUROC of 0.85 Sensitivity and specificity of 75% and 81% |
| Liu et al.29 | 2020 | Retrospective | Proprietary | 878 | Delineated by two residents reviewed by an experienced chest radiologist |
AUC of 0.93 Sensitivity and specificity of 94.6% and 76.5% |
| Huang et al.28 | 2020 | Retrospective | Proprietary | 1837 | Board-certified radiologist |
AUROC of 0.95 Sensitivity and specificity of 87.3% and 90.2% |
| Weikert et al.30 | 2019 | Retrospective | Proprietary | 29,465 | Board-certified radiologist | Sensitivity and specificity of 92.7% and 95.5% |
| Yang et al.40 | 2019 | Retrospective | Proprietary + PE challenge data | 129 | Board-certified radiologist | Sensitivity of 75.4% at two false positives per volume |
| Rajan et al. (IBM)41 | 2019 | Retrospective | Proprietary | 2420 | Board-certified radiologists | AUC of 0.94 |
| Tajbakhsh et al.26 | 2019 | Retrospective | Proprietary + PE challenge data | 121 | N/A | Sensitivity of 83% at two false positives per volume |