Publication

AI and Radiomics Streamline Risk Assessment in Thoracic Trauma

June 10, 2025

🧠 Introduction

Blunt chest trauma is among the most complex injuries to manage in emergency care. Despite established clinical protocols, there's a critical gap: no reliable tools exist to predict how a trauma case might evolve. Our research uses AI and radiomics to fill that gap—enabling more accurate, data-driven decisions in patient care.

🔍 Research Summary

Led by our team at APMAD, this study focused on optimizing the management of patients with blunt thoracic trauma. We analyzed a dataset of 212 cases from a trauma registry, applying radiomics feature extraction using the Lung CT Analyzer and PyRadiomics. Machine learning models were trained on three types of data: pre-hospital (ambulance), in-hospital (clinical and lab), and radiomics (CT-based imaging features).

Key findings include:

  • Radiomics features from lung CT scans were the top-performing predictors, achieving a sensitivity of 96.3% for forecasting the need for intensive care.
  • Models integrating ambulance, clinical, and imaging data significantly outperformed those based on any single data source.
  • Features such as age, vital signs, AIS scores for various body regions, and hemorrhage markers contributed strongly to predicting hospital length of stay (LOS).

We found that first-order kurtosis and features from the right lung—which has a larger volume—had the highest predictive value among radiomics metrics.

🌍 Broader Impact

Our study underscores the power of AI and radiomics in elevating trauma care, providing actionable insights where traditional clinical assessments fall short. By anticipating the trajectory of a trauma case early on, clinicians can tailor interventions, improve outcomes, and reduce ICU congestion.

📎 Reference 

@article{Hefny2024,
  title     = {Streamlining management in thoracic trauma: radiomics- and AI-based assessment of patient risks},
  journal   = {Frontiers in Surgery},
  volume    = {Volume 11 - 2024},
  year      = {2024},
  issn      = {2296-875X},
  doi       = {https://doi.org/10.3389/fsurg.2024.1462692},
  url       = {https://www.frontiersin.org/journals/surgery/articles/10.3389/fsurg.2024.1462692},
  author    = {Hefny, Ashraf F. and Almansoori, Taleb M. and Smetanina, Darya and Morozova, Daria and Voitetskii, Roman and Das, Karuna M. and Kashapov, Aidar and Mansour, Nirmin A. and Fathi, Mai A. and Khogali, Mohammed and Ljubisavljevic, Milos and Statsenko, Yauhen}
}

Leave a Reply

Your email address will not be published. Required fields are marked *