Uncertainty Quantification-Incorporated Ensemble Model for Personalized Prediction of Severe Radiotherapy-Induced Immunosuppression Among Esophageal Cancer Patients
Abstract
Severe radiation-induced lymphopenia (RIL), specifically Grade 4 RIL (G4RIL), is a frequent immune toxicity and a strong predictor of poor survival for esophageal cancer patients undergoing concurrent chemoradiation therapy. This study developed an uncertainty-incorporated ensemble machine learning model to predict Absolute Lymphocyte Count (ALC) nadir using baseline clinical and dosimetric variables, supporting treatment decisions such as proton therapy selection and dose distribution optimization. A cohort of 1,491 esophageal cancer patients treated with photon or proton therapy between July 2001 and April 2022 was analyzed. A weighted ensemble model was constructed using clinical features and dosimetric variables, including the composite dosimetric score (CDS) derived from radiation dose distributions. To quantify individual-level predictive uncertainty, Cross-Residual Conformal Prediction (CRCP) was developed and evaluated against standard conformal prediction methods. Proton therapy patients maintained a significantly higher mean ALC nadir than photon patients (0.33 vs. 0.25 K/uL). The ensemble model achieved a mean absolute error of 0.0926 K/uL for ALC nadir prediction and an ROC-AUC of 0.78 for G4RIL classification. Baseline ALC, Planning Target Volume, and CDS were the most influential predictors. CRCP achieved competitive prediction interval widths with mathematically guaranteed coverage, offering a favorable balance between interval width and coverage accuracy. This framework enables risk-aware, personalized clinical decision-making in radiation oncology.
Related
- Representation-aware conformal prediction for reliable uncertainty quantification in drug-target affinity
- Uncertainty Quantification in Stochastic Dynamical Gene Regulatory Networks
- Uncertainty Quantification in Cardiac Model Personalisation from Ultrafast Ultrasound
- Bayesian Uncertainty Quantification of Breast-Cancer Pathology AI Agreement under Reader- and Annotation-Condition-Dependent Reference Labels
- An MMP-14/STAT3 feedback loop upregulates PD-L1 to drive immunosuppression and tumorigenesis in bladder cancer
BibTeX
@misc{gao2026uncertainty,
title = {{Uncertainty Quantification-Incorporated Ensemble Model for Personalized Prediction of Severe Radiotherapy-Induced Immunosuppression Among Esophageal Cancer Patients}},
author = {Baode Gao and Yan Chu and Zongsheng Hu and Madison E. Grayson and Peter S. N. van Rossum and Clemens Grassberger and Heiko Enderling and Harald Paganetti and Ibrahim Chamseddine and Steven H. Lin and Brian P. Hobbs and Radhe Mohan and Yiqing Chen},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.25139},
eprint = {2609.25139},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.25139},
}
RIS
TY - UNPB
TI - Uncertainty Quantification-Incorporated Ensemble Model for Personalized Prediction of Severe Radiotherapy-Induced Immunosuppression Among Esophageal Cancer Patients
AU - Baode Gao
AU - Yan Chu
AU - Zongsheng Hu
AU - Madison E. Grayson
AU - Peter S. N. van Rossum
AU - Clemens Grassberger
AU - Heiko Enderling
AU - Harald Paganetti
AU - Ibrahim Chamseddine
AU - Steven H. Lin
AU - Brian P. Hobbs
AU - Radhe Mohan
AU - Yiqing Chen
PY - 2026
DA - 2026/09/21
PB - arXiv
DO - 10.48550/arxiv.2609.25139
UR - https://arxiv.org/abs/2609.25139
AB - Severe radiation-induced lymphopenia (RIL), specifically Grade 4 RIL (G4RIL), is a frequent immune toxicity and a strong predictor of poor survival for esophageal cancer patients undergoing concurrent chemoradiation therapy. This study developed an uncertainty-incorporated ensemble machine learning model to predict Absolute Lymphocyte Count (ALC) nadir using baseline clinical and dosimetric variables, supporting treatment decisions such as proton therapy selection and dose distribution optimization. A cohort of 1,491 esophageal cancer patients treated with photon or proton therapy between July 2001 and April 2022 was analyzed. A weighted ensemble model was constructed using clinical features and dosimetric variables, including the composite dosimetric score (CDS) derived from radiation dose distributions. To quantify individual-level predictive uncertainty, Cross-Residual Conformal Prediction (CRCP) was developed and evaluated against standard conformal prediction methods. Proton therapy patients maintained a significantly higher mean ALC nadir than photon patients (0.33 vs. 0.25 K/uL). The ensemble model achieved a mean absolute error of 0.0926 K/uL for ALC nadir prediction and an ROC-AUC of 0.78 for G4RIL classification. Baseline ALC, Planning Target Volume, and CDS were the most influential predictors. CRCP achieved competitive prediction interval widths with mathematically guaranteed coverage, offering a favorable balance between interval width and coverage accuracy. This framework enables risk-aware, personalized clinical decision-making in radiation oncology.
ER -