Catching the most common human cancer before it becomes visible to the eye — and doing so non-invasively — represents a meaningful shift in dermatologic screening capability. For the roughly one in ten high-risk patients who harbor an invisible facial skin cancer at any given moment, earlier detection directly translates to less disfiguring surgery and better cosmetic outcomes, particularly in anatomically sensitive regions like the nose and eyelids.
This prospective feasibility study enrolled 150 consecutive inpatients at a German university dermatology center, all carrying at least two established basal cell carcinoma (BCC) risk factors. Researchers systematically scanned inconspicuous facial skin — areas showing no macroscopic lesion — using line-field confocal optical coherence tomography (LC-OCT) paired with an AI image-recognition algorithm. Of 18 sites flagged by the AI-LC-OCT system, 15 were histologically confirmed as BCC, yielding a positive predictive value of approximately 83%. Across the cohort, 17 subclinical BCCs were identified in 14 patients, an overall subclinical prevalence of 9.3% among this high-risk group.
LC-OCT generates near-histological cross-sectional images of skin at depths up to 500 µm without tissue removal, and its pairing with AI classification is a logical extension of a technology already validated for visible lesions. What makes this study notable is the systematic screening application to clinically normal-appearing skin — a use case that has not been well-studied. That said, several important caveats apply. The study is small, single-center, and cross-sectional, meaning sensitivity (the false-negative rate) cannot be fully characterized from this dataset alone. The patient population — hospitalized, elderly, predominantly male, mean age 73 — limits generalizability to outpatient or younger cohorts. The AI model's performance on more diverse skin types also remains untested. Overall, this is an incremental but directionally important proof-of-concept: AI-assisted optical imaging may eventually enable routine, non-biopsy screening of high-risk skin, though larger multicenter validation is needed before clinical adoption.