Table 1.
Patient characteristics in the curated evaluation dataset.
| Characteristics | Num. Cases (%) |
|---|---|
| No. of cases | 5420 |
| No. of images included in study | 14,303 |
| No. of patients included in study | 4180 |
| Female (%) | 2980 (55.0%) |
| Age, median (25th, 75th percentiles) | 55 (39, 67) |
| Race and ethnicitya | |
| White, non-Hispanic (%) | 698 (32.2%) |
| Asian and Pacific Islander, non-Hispanic (%) | 261 (12.0%) |
| Hispanic (%) | 867 (40.0%) |
| Black, non-Hispanic (%) | 291 (13.4%) |
| Other/Not specified (%) | 51 (2.4%) |
| Estimated Fitzpatrick skin types | |
| I or II (%) | 2621 (48.4%) |
| III or IV (%) | 2026 (37.4%) |
| V or VI (%) | 262 (4.8%) |
| N/A (%) | 511 (9.4%) |
| Skin conditions based on reference diagnosis | Num. Cases (%) | Skin conditions based on reference diagnosis | Num. Cases (%) |
|---|---|---|---|
| Acne | 135 (2.5%) | Psoriasis | 93 (1.7%) |
| Actinic keratosis | 357 (6.6%) | SCC/SCCISb | 493 (9.1%) |
| Allergic contact dermatitis | 57 (1.1%) | SK/ISK | 377 (7.0%) |
| Alopecia areata | 46 (0.8%) | Scar condition | 61 (1.1%) |
| Androgenetic alopecia | 45 (0.8%) | Seborrheic dermatitis | 46 (0.8%) |
| Basal cell carcinomab | 518 (9.6%) | Skin tag | 37 (0.7%) |
| Cyst | 111 (2.0%) | Stasis dermatitis | 36 (0.7%) |
| Eczema | 159 (2.9%) | Tinea | 35 (0.6%) |
| Folliculitis | 53 (1.0%) | Tinea versicolor | 31 (0.6%) |
| Hidradenitis | 44 (0.8%) | Urticaria | 17 (0.3%) |
| Lentigo | 152 (2.8%) | Verruca vulgaris | 44 (0.8%) |
| Melanocytic nevus | 563 (10.4%) | Vitiligo | 41 (0.8%) |
| Melanomab | 195 (3.6%) | Other condition | 1636 (30.2%) |
| Post-inflammatory hyperpigmentation | 38 (0.7%) |
Enrichment was performed to mitigate skew towards common demographics and dermatologic conditions, and additionally to include all available cases with race/ethnicity information. The table includes 26 common skin conditions, representing 80% of cases seen in primary care,20 and an additional category grouping other conditions.
a
Race/ethnicity information was unavailable for the skin cancer dataset. Case counts and percentages for race/ethnicity therefore reflect only cases from the tele-dermatology dataset where race/ethnicity was generally available.
b
Skin cancers.