Prediction of Tattoo Removal Sessions With Picosecond Lasers: Development and Validation of a Novel Clinical Scale With Machine Learning Techniques

Le résumé fourni par la source

ABSTRACT Background Accurate prediction of the number of sessions required for laser tattoo removal is essential for effective treatment planning and patient counseling. The Kirby–Desai (KD) scale, developed for Q‐switched nanosecond lasers, overestimates treatment sessions when applied to modern picosecond laser technology. Aims To develop and validate a novel predictive scoring system specifically calibrated for picosecond laser tattoo removal and to compare its predictive accuracy with the KD scale. Methods A retrospective cohort of 545 patients treated with picosecond laser tattoo removal between 2019 and 2024 was analyzed. Tattoo characteristics were evaluated using both the KD scale and the newly developed (PT) scale, which incorporates six parameters: tattoo type, pigment density, anatomical location, ink color, scarring, and layering. Multiple regression and machine learning models were trained separately on KD‐ and PTbased variables. Model performance was assessed using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination ( R 2 ) on an independent test set. Model interpretability was evaluated using SHapley Additive exPlanations (SHAP). Results PT‐based models consistently outperformed KD–based models. The PT Ridge regression model achieved the best predictive performance (MAE = 0.58, RMSE = 0.64, R 2 = 0.73), significantly improving accuracy compared with KD‐based approaches ( p < 0.001). SHAP analysis identified tattoo type and pigment density as the most influential predictors. Conclusions The PT scale provides a more accurate and clinically relevant prediction of treatment sessions for picosecond laser tattoo removal than the KD scale, offering a practical and reliable tool for contemporary clinical practice.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.