Sign in Subscribe
Artificial neural networks

Neural Networks Enhance Burn Wound Healing Predictions

Strategic Insights

— 1 min read

Burn injuries trigger substantial inflammation, complicating wound healing and potentially leading to severe systemic complications. Understanding the immune response to burns is crucial for improving treatment. Researchers from the University of Amsterdam have developed a novel approach using neural networks to predict cytokine concentrations over time and space, which can aid in burn wound healing. This study integrates neural networks as surrogate models to approximate and forecast agent-based model simulation results, offering a more efficient and effective solution for predicting cytokine concentrations.

Key Takeaways:

  • The research, supported by the Dutch Burns Foundation (DBF) - Health-Holland, aims to improve treatment outcomes for burn injuries by understanding the immune response.
  • The University of Amsterdam team developed a baseline agent-based model using CompuCell3D software to simulate the innate immune response and generate extensive cytokine concentration data.
  • The team processed and prepared the data for neural network training, involving data cleaning, transformation into suitable formats, and a time-series-aware train-test split.
  • Various neural network architectures, including convolutional neural networks (CNNs), long short-term memory (LSTM) neural networks, attention mechanisms, and physics-informed neural networks (PINNs), were implemented and assessed.
  • The research concluded that STA-LSTM generally performs best across statistical metrics.
  • The study has been peer-reviewed and published in the Journal of Computational Science.
  • The research team included Vivek M. Sheraton, Ioannis Papapanagiotou, Roland V. Bumbuc, Valeria Krzhizhanovskaya, and H. Ibrahim Korkmaz.

Statistics:

  • The study was supported by the Dutch Burns Foundation (DBF) - Health-Holland.
  • The research was published in the Journal of Computational Science, Volume 89, 2025.
  • The team processed and prepared over 10,000 data points for neural network training.
  • Statistical metrics, including Mean Squared Error, R-squared, and Mean Absolute Percentage Error, were used to evaluate the performance of the neural network models.
  • The study concluded that STA-LSTM generally performs best across statistical metrics, with a Mean Squared Error of 0.012 and an R-squared value of 0.95.

Sources:

  • "From Simulations To Surrogates: Neural Networks Enhancing Burn Wound Healing Predictions." Journal of Computational Science, 2025; 89.
  • "Investigators from University of Amsterdam Zero in on Cytokines." Health & Medicine Week. July 11, 2025; p 2344.
  • Dutch Burns Foundation (DBF) - Health-Holland
  • Elsevier - www.elsevier.com
  • Journal of Computational Science - www.journals.elsevier.com/journal-of-computational-science/

Read more