Table 4.
Summary estimate of pooled performance of artificial intelligence in image-based lung cancer prognosis prediction
| Subgroup | No. of studies (datasets) | Tau2 | Sensitivity | Specificity | P valueb | ||||
|---|---|---|---|---|---|---|---|---|---|
| Sensitivity | P valuea | I2 (95%CI) | Specificity | P valuea | I2 (95%CI) | ||||
| Overall | 58 (78) | 0.022 | 0.83 (0.81–0.86) | < 0.05 | 64.97 (56.60–73.35) | 0.83 (0.80–0.86) | < 0.05 | 76.96 (72.04–81.87) | |
| Cohort | 0.0237 | ||||||||
| Internal validation cohort | (52) | 0.0139 | 0.85 (0.82–0.88) | < 0.05 | 70.12 (61.72–78.51) | 0.84 (0.81–0.87) | < 0.05 | 73.18 (65.87–80.50) | |
| External validation cohort | (26) | 0.0560 | 0.79 (0.75–0.82) | < 0.05 | 36.62 (6.51–66.72) | 0.81 (0.75–0.85) | < 0.05 | 78.45 (70.57–86.32) | |
| Algorithm | 0.9761 | ||||||||
| Machine learning | 45 (57) | 0.0083 | 0.84 (0.81–0.86) | < 0.05 | 60.56 (49.18–71.95) | 0.82 (0.79–0.85) | < 0.05 | 66.31 (56.98–75.65) | |
| Deep learning | 13 (21) | 0.0669 | 0.82 (0.77–0.86) | < 0.05 | 74.29 (63.29–85.28) | 0.87 (0.80–0.92) | < 0.05 | 88.72 (84.89–92.56) | |
| Imaging variable only | 0.0876 | ||||||||
| Yes | 32 (46) | 0.0229 | 0.84 (0.80–0.87) | < 0.05 | 61.30 (55.90–76.70) | 0.86 (0.82–0.89) | < 0.05 | 76.86 (70.40–83.31) | |
| No | 26 (32) | 0.0231 | 0.83 (0.79–0.86) | < 0.05 | 61.09 (46.13–76.04) | 0.79 (0.75–0.83) | < 0.05 | 73.46 (64.23–82.69) | |
| Nodule segmentation | 0.6671 | ||||||||
| Manual | 30 (35) | 0.0125 | 0.83 (0.79–0.87) | < 0.05 | 71.34 (61.61–81.06) | 0.83 (0.78–0.86) | < 0.05 | 71.04 (61.19–80.89) | |
| Algorithm | 28 (43) | 0.0332 | 0.83 (0.80–0.85) | < 0.05 | 59.36 (45.74–72.98) | 0.84 (0.80–0.87) | < 0.05 | 80.52 (75.16–85.88) | |
| Image quality control | 0.2975 | ||||||||
| Yes | 13 (15) | 0.0196 | 0.86 (0.80–0.91) | < 0.05 | 71.47 (56.58–86.36) | 0.85 (0.79–0.90) | < 0.05 | 61.31 (39.57–83.06) | |
| No | 45 (63) | 0.023 | 0.83 (0.80–0.85) | < 0.05 | 62.19 (51.92–72.46) | 0.83 (0.79–0.85) | < 0.05 | 78.64 (73.67–83.61) | |
| Imaging modality | 0.7405 | ||||||||
| CT | 40 (57) | 0.0300 | 0.84 (0.81–0.86) | < 0.05 | 63.36 (52.98–73.74) | 0.83 (0.79–0.86) | < 0.05 | 77.46 (71.87–83.06) | |
| CECT | 7 (9) | 0.0000 | 0.83 (0.75–0.89) | < 0.05 | 66.60 (43.05–90.15) | 0.90 (0.75–0.97) | < 0.05 | 85.03 (76.42–93.65) | |
| PET/CT | 11 (12) | 0.0184 | 0.82 (0.75–0.88) | < 0.05 | 75.36 (61.45–89.26) | 0.83 (0.78–0.87) | < 0.05 | 62.78 (39.56–85.99) | |
aP-Value for heterogeneity within each subgroup.
bP-Value for heterogeneity between subgroups with multivariable meta-regression analysis.