Table 2 Accuracy of rule set 9 applied to AI system segmentation outputs for determining disease activity over subgroups defined by age, sex, ethnicity and diabetic status.
Characteristic | n | TP | TN | FP | FN | Accuracy | 95% CI | |
|---|---|---|---|---|---|---|---|---|
Age | 58–75 | 130 | 28 | 78 | 20 | 4 | 81.5% | 74.9–88.2% |
75–82 | 130 | 30 | 73 | 23 | 4 | 79.2% | 72.3–86.2% | |
82–86 | 131 | 27 | 76 | 24 | 4 | 78.6% | 71.6–85.6% | |
86–98 | 130 | 30 | 80 | 17 | 3 | 84.6% | 78.4–90.8% | |
Sex | Male | 194 | 41 | 114 | 32 | 7 | 79.9% | 74.3–85.5% |
Female | 327 | 74 | 193 | 52 | 8 | 81.7% | 77.5–85.8% | |
Diabetic | Yes | 59 | 11 | 33 | 14 | 1 | 74.6% | 63.5–85.7% |
No/not known | 462 | 104 | 274 | 70 | 14 | 81.8% | 78.3–85.3% | |
Ethnicity | White | 414 | 96 | 235 | 72 | 11 | 80.0% | 76.1–83.8% |
Asian | 31 | 5 | 22 | 3 | 1 | 87.1% | 75.4–98.9% | |
Other | 39 | 4 | 27 | 6 | 2 | 79.5% | 66.8–92.2% | |
Unknown | 38 | 10 | 23 | 3 | 2 | 86.8% | 76.1–97.6% |
