Dermoscopic Differentiation of Actinic Keratosis and Bowen’s Disease: A Model for Challenging Facial High‐Grade Lesions

Background/Objectives This study aimed to comprehensively analyze the dermoscopic characteristics of Bowen’s disease (BD) across different anatomical locations and actinic keratosis (AK) by pathological grade, with a specific focus on developing a diagnostic model to differentiate challenging facial high‐grade AK from BD. Methods A prospective study was conducted involving patients with histopathologically confirmed AK or BD who underwent standardized dermoscopic imaging. AK lesions were systematically evaluated using three classification systems: Röwert‐Huber (histopathological), dermoscopic grading, and Olsen (clinical). Statistical analyses included interobserver agreement, correlation tests, and logistic regression to identify discriminative features and build a diagnostic model. Results A total of 113 patients were included, comprising 78 AK cases (113 lesions) and 55 BD cases (65 lesions). The cohort included 31 AK I, 37 AK II, and 45 AK III lesions according to Röwert‐Huber pathological grading. Analysis of lesion distribution revealed significantly more facial than scalp lesions across AK grades and BD ( p = 0.043). Dermoscopic analysis revealed distinct feature distributions between AK and BD: red pseudonetwork, strawberry pattern, perifollicular white halo, follicular plugs, rosette sign, and brown/gray pseudonetwork were characteristic of AK, whereas erosion, blood crust, yellow crust, irregular linear/serpentine vessels, and glomerular vessels were significantly associated with BD (all p < 0.05). For differentiating facial Grade AK III from BD, multivariate analysis identified “irregular linear/serpentine vessels” as an independent predictor for BD (OR = 12.072), while “red pseudonetwork” (OR = 0.188) and “follicular keratotic plugs” (OR = 0.174) significantly indicated AK III. A diagnostic model incorporating these three features achieved an AUC of 0.88 (95% CI: 0.81–0.95), with 93.3% sensitivity and 71.1% specificity. Conclusions Our findings establish a robust dermoscopic model for noninvasive discrimination between high‐grade AK and BD, offering a practical tool for clinical decision‐making in managing facial lesions.