A predictive model for clinical outcomes in human acute radiation syndrome

Read full text Summarise with VIRAT
Authors: V.I. Pustovoyt, A.S. Umnikov, Aleksey Valeryevich Bogomolov
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Acute radiation syndrome remains one of the most clinically complex forms of human radiation injury, necessitating accurate and reproducible tools for forecasting clinical outcomes. Given the ethical and practical limitations associated with conducting randomized clinical trials in radiation medicine, predictive models based on retrospective data acquire particular significance. The aim of this study was to ensure an acceptable level of accuracy in forecasting clinical outcomes of human ARS, to standardize diagnostic procedures, and to justify personalized therapeutic strategies under conditions of limited time and resources through the development and validation of a prognostic model that integrates individual radiation dose burdens, clinical and laboratory indicators, and disease dynamics. The model was constructed using an international dataset derived from a retrospective analysis of 1441 clinical cases of acute radiation syndrome accumulated over the past 70 years, utilizing linear discriminant analysis. The forecasted clinical outcome – favorable or unfavorable – is based on six clinico-radiobiological parameters: cytogenetic dose, severity of the syndrome, radionuclide incorporation, time elapsed since exposure, duration of exposure, and patient body weight. The model demonstrated acceptable accuracy: with an overall accuracy of 94,8%, the predictive accuracy for favorable outcomes was 97,4%, and for unfavorable outcomes – 80,3%. The high performance of the model was corroborated by the results of the classification matrix and ROC analysis. This model represents the first documented application of linear discriminant analysis for blind classification of ARS clinical outcomes in humans based on clinico-radiobiological parameters, thus offering reproducibility, interpretability, and potential for automation in systems of medical triage, diagnosis, and emergency response during radiation incidents.

Fields of Study & Subjects

Effects of Radiation Exposure Radiation Dose and Imaging Radioactive contamination and transfer

Publication details

Journal / source
Radiation and Risk Bulletin of the National Radiation and Epidemiological Registry
Volume
35
Issue
1
Pages
117–126
Publication date
1 January 2026
Language
EN