Radiomic and dosiomic machine learning models for predicting radiation-induced oral mucositis in head and neck cancer: a systematic review.
Lukas Mendes de Abreu, Raquel Molina Sanches, Alana Crescencio de Oliveira et al.
✦ AI-curated · Sources linked
In 30 seconds
This PRISMA-guided systematic review assessed machine-learning models using radiomic and/or dosiomic data to predict radiation-induced oral mucositis in patients with head and neck cancer. Seven studies involving 2,311 patients were included; all models used dosimetric predictors, with some adding other imaging features or oral health indices. Reported discrimination ranged from AUC 0.70 to 0.96, while external validations showed reduced performance (AUC ~ 0.65). The review identified high risk of bias, particularly in analysis, and concluded that these models remain exploratory rather than ready for routine clinical use.
Key findings
- Seven studies involving 2,311 patients were included, and all models incorporated dosimetric predictors.
- Reported AUC ranged from 0.70 to 0.96; external validations showed reduced performance (AUC ~ 0.65).
- High risk of bias was mainly identified in analysis domains, associated with limited sample sizes and incomplete calibration.
Why it matters
Predicting oral mucositis could help characterize risk during radiotherapy, but model performance in development studies may not translate to other settings. The reduced discrimination in external validation and concerns about calibration and bias indicate that reported accuracy should be interpreted cautiously before these tools inform clinical decisions.
What to keep in mind
The review found limited sample sizes and incomplete calibration, with high risk of bias mainly in analysis domains; external validation performance was reduced.