Improved detection of lentigo maligna with AI-assisted dermoscopy: A reader study in facial pigmented lesions.

BACKGROUND

Differentiating lentigo maligna (LM) from benign facial pigmented lesions remains difficult due to substantial clinical and dermoscopic overlap, particularly on chronically sun-damaged skin, a setting underrepresented in existing artificial intelligence (AI) studies.

OBJECTIVE

To develop and evaluate a deep learning-based model for facial pigmented lesions and assess its impact on dermatology resident diagnostic performance.

METHODS

In this retrospective study, 722 lesions (894 dermoscopic images) were analyzed (LM: 190; pigmented actinic keratosis: 230; solar lentigo/seborrheic keratosis: 302). Twenty percent of lesions were reserved for testing; the remainder underwent five-fold stratified cross-validation. An Xception-based convolutional neural network was trained for binary and 3-class classification. A reader study with 26 residents compared diagnostic accuracy before and after AI assistance.

RESULTS

The model achieved a mean accuracy of 84.2% ± 2.5%, sensitivity of 90.0% ± 11.5%, and specificity of 81.9% ± 1.2%. Resident accuracy improved from 64.9% to 74.0% with AI support (P < .0001), with the largest gain observed in LM detection (+16.5%).

LIMITATIONS: Retrospective design, lack of multimodal clinical data, and resident-only reader study.

CONCLUSION

AI-assisted dermoscopy improves diagnostic performance in a challenging facial lesion setting, particularly for LM, supporting its role as an adjunct tool in clinical decision-making and dermatology training.