Table 6.
Concordance between a software-identified match within the top five closest matches and whether two of three clinical geneticists would definitely have considered this diagnosis based on the image alone
| Syndrome | Frequency | Correctly classified by neither | Correctly classified by both | Correctly classified by software, but not clinicians | Correctly classified by clinicians, but not software | McNemars Chi-square | p-value | Kappa |
|---|---|---|---|---|---|---|---|---|
| Williams | 183 | 2 | 103 | 73 | 5 | 59.28 | <.00001 | −0.02 |
| Rubinstein-Taybi | 155 | 4 | 94 | 55 | 2 | 49.28 | <.00001 | 0.06 |
| Floating Harbor | 61 | 1 | 43 | 15 | 2 | 9.94 | 0.00162 | 0.02 |
| Coffin Lowry | 154 | 15 | 59 | 74 | 6 | 57.80 | <.00001 | 0.07 |
| Kabuki | 195 | 2 | 148 | 39 | 6 | 24.20 | <.00001 | 0.01 |
| Smith Magenis | 124 | 2 | 56 | 66 | 0 | 66.00 | <.00001 | 0.03 |
| PACS1 | 39 | 20 | 2 | 17 | 0 | 17.00 | 0.00004 | 0.11 |
| Kleefstra | 128 | 18 | 11 | 97 | 2 | 91.16 | <.00001 | 0.00 |
| Koolan-de Vries | 120 | 17 | 54 | 42 | 7 | 25.00 | <.00001 | 0.18 |
| Cornelia de Lange | 441 | 10 | 311 | 107 | 13 | 73.63 | <.00001 | 0.06 |