Background/Objectives: Melanoma is the most lethal cutaneous neoplasm, with Breslow thickness being a key prognostic factor. This retrospective cohort study aimed to assess the impact of screening frequency and diagnostic methods on tumour stage at diagnosis and to explore implications for risk-adapted strategies. Methods: Between 2017 and 2024, 475 cases of melanoma were diagnosed in 397 patients. Screening frequency, diagnostic method, and patient risk were analyzed in relation to tumour stage. Results: Compared with first-visit cases, patients who underwent screening within two years prior to diagnosis were more often diagnosed with melanoma in situ (32.6% vs. 44–51%; p < 0.05) and had thinner invasive tumours (0.68–0.73 mm vs. 1.8 mm; p ≤ 0.001), though no differences were seen between screening frequencies. Full-body examination was associated with more in situ melanomas (46% vs. 34%; p = 0.016) and thinner invasive tumours (0.92 vs. 2.05 mm; p = 0.2) compared with lesion-directed screening, but this effect disappeared after excluding first-visit cases. Invasive melanomas diagnosed by mole mapping were significantly thinner than by dermoscopy (0.55 vs. 1.07; p = 0.035). In high-risk patients, tumour thickness decreased with more frequent visits (0.905 mm without screening vs. 0.40–0.55 mm with ≥1 visit; p = 0.001). Moreover, mole mapping identified thinner melanomas in the high-risk group compared with dermoscopy (0.47 vs. 0.60 mm; p = 0.02). Conclusions: Screening is associated with thinner melanomas and more in situ diagnoses. Digital mole mapping offers additional benefits, with high-risk patients profiting most, while low-risk individuals could be managed with less resource-intensive approaches. These findings support risk-adapted screening strategies focusing on intensive, digitally supported modalities for high-risk groups.

In order to improve the detection and classification of malignant melanoma, this paper describes an image-based method that can achieve AUROC values of up to 0.78 without additional clinical information. Furthermore, the importance of the domain gap between two different image sources is considered, as it is important to create usability independent of hardware components such as the high-resolution scanner used. Since for the application of machine learning methods, alterations of scanner-specific properties such as brightness, contrast or sharpness can have strong (negative) effects on the quality of the prediction methods, two ways to overcome this domain gap are discussed in this paper.