Predicting acute radiation dermatitis in breast cancer patients using artificial intelligence: a systematic review.
- 1Department of Radiotherapy Physics & Technology, West China Hospital, Sichuan University, Chengdu, China.
- 2Health Management Center, General Practice Medical Center, West China Hospital, Sichuan University, Chengdu, China.
Abstract
Acute radiation dermatitis (ARD) is a common toxicity of breast cancer radiotherapy that affects quality of life. Artificial intelligence (AI)-based models have been increasingly developed to predict ARD, but their methodological quality and clinical readiness remain unclear. This study aimed to systematically evaluate AI-based prediction models for ARD. A systematic search of PubMed, Embase (Ovid), Web of Science, and MEDLINE (Ovid) was conducted for studies published up to December 1, 2025. Studies were eligible if they developed and/or validated AI-based models for predicting ARD in patients with breast cancer receiving radiotherapy. Data were extracted using the CHARMS framework. Model quality, applicability, and risk of bias were assessed using the Prediction Model Risk of Bias Assessment Tool plus Artificial Intelligence (PROBAST + AI) tool. Given the substantial heterogeneity among studies, a qualitative synthesis was performed. Nineteen studies involving 8,506 patients were included, with sample sizes ranging from 20 to 2,058. Most studies were single-center (16/19), and only three were multicenter. Validation area under the curve (AUC) values ranged from 0.66 to 0.99. The models incorporated clinical, dosimetric, radiomics, and/or dosiomics features, with 5-154 predictors in the final models. Calibration and decision curve analysis were rarely reported, and no study performed external validation. Most studies had a high or unclear risk of bias according to PROBAST + AI, mainly related to small sample sizes, incomplete reporting, and lack of external validation. Although current AI-based models for predicting ARD in breast cancer radiotherapy demonstrate promising discriminatory performance, their methodological limitations and lack of external validation limit their readiness for routine clinical implementation. Future studies with more rigorous designs, standardized outcome assessment, transparent reporting, and independent external validation are required to facilitate clinical translation.
Related
- Artificial intelligence-enhanced electrocardiography for predicting mortality and major adverse cardiovascular events: A systematic review and narrative synthesis
- The role of artificial intelligence in predicting late-onset intrauterine growth restriction in singleton pregnancies: a systematic review and meta-analysis
- Educational Artificial Intelligence Software to Support Assessment of Atopic Dermatitis Severity
- Atopic Dermatitis in the Era of Exposome and Artificial Intelligence
- Predicting follow-up cognition using estimated structural and functional connectivity networks and artificial intelligence in multiple sclerosis
BibTeX
@article{yu2026predicting,
title = {{Predicting acute radiation dermatitis in breast cancer patients using artificial intelligence: a systematic review.}},
author = {Hang Yu and Pan Gong and Xiangbin Zhang and Guijun Xue and Renming Zhong},
journal = {Quantitative imaging in medicine and surgery},
year = {2026},
month = oct,
doi = {10.21037/qims-2026-1245},
url = {https://europepmc.org/article/MED/42997736},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Predicting acute radiation dermatitis in breast cancer patients using artificial intelligence: a systematic review.
AU - Hang Yu
AU - Pan Gong
AU - Xiangbin Zhang
AU - Guijun Xue
AU - Renming Zhong
PY - 2026
DA - 2026/10/19
JO - Quantitative imaging in medicine and surgery
DO - 10.21037/qims-2026-1245
UR - https://europepmc.org/article/MED/42997736
AB - Acute radiation dermatitis (ARD) is a common toxicity of breast cancer radiotherapy that affects quality of life. Artificial intelligence (AI)-based models have been increasingly developed to predict ARD, but their methodological quality and clinical readiness remain unclear. This study aimed to systematically evaluate AI-based prediction models for ARD. A systematic search of PubMed, Embase (Ovid), Web of Science, and MEDLINE (Ovid) was conducted for studies published up to December 1, 2025. Studies were eligible if they developed and/or validated AI-based models for predicting ARD in patients with breast cancer receiving radiotherapy. Data were extracted using the CHARMS framework. Model quality, applicability, and risk of bias were assessed using the Prediction Model Risk of Bias Assessment Tool plus Artificial Intelligence (PROBAST + AI) tool. Given the substantial heterogeneity among studies, a qualitative synthesis was performed. Nineteen studies involving 8,506 patients were included, with sample sizes ranging from 20 to 2,058. Most studies were single-center (16/19), and only three were multicenter. Validation area under the curve (AUC) values ranged from 0.66 to 0.99. The models incorporated clinical, dosimetric, radiomics, and/or dosiomics features, with 5-154 predictors in the final models. Calibration and decision curve analysis were rarely reported, and no study performed external validation. Most studies had a high or unclear risk of bias according to PROBAST + AI, mainly related to small sample sizes, incomplete reporting, and lack of external validation. Although current AI-based models for predicting ARD in breast cancer radiotherapy demonstrate promising discriminatory performance, their methodological limitations and lack of external validation limit their readiness for routine clinical implementation. Future studies with more rigorous designs, standardized outcome assessment, transparent reporting, and independent external validation are required to facilitate clinical translation.
SN - 2223-4292
SN - 2223-4306
ER -