Predicting shock-induced cavitation using machine learning: implications for blast-injury models.

Area of Science:

• Biophysics
• Computational Biology
• Neuroscience

Background:

• Cavitation is a suspected mechanism in blast-induced traumatic brain injury (bTBI).
• Studying cavitation in vivo is challenging, necessitating reliance on numerical simulations.
• Validating these simulations with experimental data is crucial for accurate bTBI research.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in predicting shock-induced cavitation.
• To compare the predictive performance of k-nearest neighbor (kNN) and support vector machine (SVM) models.
• To demonstrate the potential of machine learning in advancing blast injury research.

Main Methods:

• Developed and trained kNN and SVM machine learning models.
• Utilized experimental data from a three-dimensional shock tube model for training and validation.
• Assessed the accuracy of the models in predicting cavitation bubble formation.

Main Results:

• Both kNN and SVM algorithms demonstrated high accuracy in predicting the number of cavitation bubbles.
• The models successfully predicted cavitation behavior based on experimental parameters like temperature.
• Machine learning models proved effective in correlating experimental and simulation data.

Conclusions:

• Machine learning offers a viable approach for studying biological cavitation and blast injury.
• This study validates the use of machine learning for predicting cavitation phenomena relevant to bTBI.
• The findings highlight the potential utility of ML in understanding and mitigating blast-induced neurological damage.