Introduction. Early diagnosis of malignant skin neoplasms is critical for prognosis, yet the shortage of oncodermatologists in Russian regions limits screening coverage. Mobile dermoscopy-based clinical decision support systems (CDSS) are a promising tool; however, proof of diagnostic efficacy and equipping primary care remain prerequisites, while interpretability of the output and standardisation of routing become critical at the stage of clinical integration.
Aim. To assess the concordance of Melanoscope AI CDSS conclusions with independent expert dermoscopic evaluation, and to evaluate the change in general practitioner (GP) conclusions after the system’s result is presented during screening sessions.
Material and methods. A single-centre prospective pilot study was conducted across four screening sessions (June 2025 – April 2026). The material comprised dermoscopic images — one most suspicious lesion per each of 176 examined patients, selected based on visual examination. Concordance of the system’s conclusions (malignant/benign) with the expert conclusion formed before reviewing the system output was assessed. Lesions considered malignant by the expert were histologically confirmed; benign cases were not morphologically verified (partial differential verification). The change in concordance of GP conclusions with the expert conclusion for the same cases, before and after presenting the system’s result, was assessed using the exact McNemar test.
Results. Agreement of the system’s binary classification with the independent expert conclusion for 176 lesions was 88.6% (95% CI 83.0–92.9), specificity 88.3% (82.5–92.7), positive predictive value 20.0% (6.8–40.7). All five histologically confirmed malignant neoplasms (3 melanomas and 2 basal cell carcinomas) were assigned by the system to the red routing zone; false negatives under the applied reference standard could not be detected, since benign cases were not histologically verified. Six patients with dysplastic naevi were placed under dynamic follow-up. Concordance of GP conclusions with the expert conclusion increased from 71.0 (63.9–77.2) to 82.4% (76.1–87.3); the gain was 11.4 pp (6.7–16.3), exact McNemar test p = 1.9·10⁻⁶.
Conclusion. All histologically confirmed malignant neoplasms were assigned by the system to the red routing zone; however, sensitivity and the safety of excluding malignancy are not established by this pilot study due to the small number of events and the absence of morphological verification of negative cases. Concordance of GP conclusions with the expert conclusion increased after the system’s decision was presented. A multicentre validation of the system’s decision concordance on a large population under screening conditions is required, as well as centralised dynamic monitoring of cases assigned by the system to the yellow routing zone.