Artificial intelligence-based automated grading and severity prediction of radiation dermatitis: a review

Le résumé fourni par la source

Radiation dermatitis (RD) is a common complication of radiotherapy, affecting up to 90% of patients. AI-based automated assessment may improve objective grading and predictive accuracy. To review AI applications in RD detection, grading, and severity prediction. This review summarizes the promising application of deep learning for automated RD grading and identifies a distinct lack of research into AI models for predicting its progression. AI shows promise in reducing interobserver variability and enabling early intervention. Challenges include dataset limitations and lack of cross-center validation. AI-based tools have potential for personalized RD management, but further multicenter validation is needed. Clinical Trial registration: ChiCTR2400082684 Comprehensive systematic review of AI applications in radiation dermatitis (RD) assessment and prediction Demonstrated 83-87% accuracy of deep learning models in automated RD severity grading Significant reduction of interobserver variability compared to traditional visual assessment methods Critical analysis of current limitations in longitudinal prediction models for RD progression Proposed multimodal fusion approach combining imaging, dosimetry, and clinical data Identified key challenges: dataset limitations, population bias, and model interpretability Future directions: multicenter validation, temporal modeling, and explainable AI frameworks

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