Resumen

BENAVIDES MORILLO, Verónica Viviana; LOPEZ CISNEROS, Ragde Carolyne  y  CARGUA RIOS, José Napoleón. A neural network-based predictive model for evaluating the efficacy of low-intensity lasers in orthodontic treatments. RCCI [online]. 2025, vol.19, n.2  Epub 25-Sep-2025. ISSN 2227-1899.

This study develops a predictive computational model based on artificial neural networks to evaluate the efficacy of low-level laser light (LLLT) in orthodontic treatments. Anonymized clinical records from the PhysioNet biomedical repository were used, complemented with experimental data collected from private clinics specializing in laser orthodontics in Ecuador. The sample included 240 patients (120 with laser, 120 without laser), recording variables such as treatment duration, reported pain levels (VAS scale), degree of inflammation, and monthly tooth displacement. The model, trained with a multilayer neural network (MLP) with 3 hidden layers and ReLU activation, was able to predict the estimated treatment time with an accuracy of 92.3%. Furthermore, the average pain in LLLT patients was 43% lower during the first 6 weeks. Cross-validation was performed using a 70/30 split, yielding an RMSE of 0.41 months in predicting total treatment time. The tool was able to identify optimal laser energy thresholds (830 nm, 20 mW for 10 s) to maximize therapeutic response without side effects. It is concluded that the computational model improves clinical planning in orthodontics and justifies the use of lasers as a support tool based on quantitative evidence. Its integration into intelligent clinical platforms is recommended for the personalization of orthodontic treatment.

Palabras clave : low-level laser; orthodontics; artificial intelligence; neural networks; predictive model.

        · resumen en Español     · texto en Español     · Español ( pdf )