Tumour samples used in the study
| Training* | Validation | |
|---|---|---|
| Cohort | ||
| #Samples | 375 | 201 |
| #Patients | 357 | 201 |
| % Malignant | 49 | 30 |
| Age (median) | 54 | 53 |
| % Females | 73 | 80 |
| Cytology | ||
| #Giemsa | 212 | 90 |
| #Diff-Quik | 95 | 21 |
| #Papanicolaou | 62 | 90 |
| #Bethesda II | 27 | 0 |
| #Bethesda III | 80 | 29 |
| #Bethesda IV | 142 | 131 |
| #Bethesda V | 77 | 41 |
| #Bethesda VI | 49 | 0 |
*Patient age was missing for 64 training samples and patient gender was missing for 10 training samples. Three training samples were created by mixing more than one slide (with different stains), two were unstained, and for one the stain was unknown.