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

While cavitation has been suspected as a mechanism of blast-induced traumatic brain injury (bTBI) for a number of years, this phenomenon remains difficult to study due to the current inability to measure cavitation in vivo. Therefore, numerical simulations are often implemented to study cavitation in the brain and surrounding fluids after blast exposure. However, these simulations need to be validated with the results from cavitation experiments. Machine learning algorithms have not generally been applied to study blast injury or biological cavitation models. However, such algorithms have concrete measures for optimization using fewer parameters than those of finite element or fluid dynamics models. Thus, machine learning algorithms are a viable option for predicting cavitation behavior from experiments and numerical simulations. This paper compares the ability of two machine learning algorithms, k-nearest neighbor (kNN) and support vector machine (SVM), to predict shock-induced cavitation behavior. The machine learning models were trained and validated with experimental data from a three-dimensional shock tube model, and it has been shown that the algorithms could predict the number of cavitation bubbles produced at a given temperature with good accuracy. This study demonstrates the potential utility of machine learning in studying shock-induced cavitation for applications in blast injury research.


Training and Testing Points using the kNN Model and Cavitation Scheme 1
Table S1.Number of training and testing points, as a function of temperature and cavitation level, for the kNN Model using cavitation scheme 1.The k = 2 (best performing value of k) data from Figures S2a and S7a

Number of Training and Testing
Points for Scheme 2 with k = 5

SUPPLEMENTARY FIGURES
The supplementary figures show the performances and corresponding confusion matrices for the k-Nearest Neighbors (kNN) models and an adapted Support Vector Machine model, using Error-Correcting Output Codes (ECOC SVM).The results from ten (10) kNN and ECOC SVM models, each using a different 70% training and 30% testing data, were averaged to obtain the cross-validation accuracies.The maximum and minimum cross-validation accuracies across all k values for each cavitation scheme are shown in a bar chart in Figure S11.Bootstrapping results for both the kNN and ECOC SVM models are shown in Figure S14 as a box-and-whisker plot.Figure S5.kNN Model Performance for k = 7 using (S5a) cavitation scheme 1, (S5b) cavitation scheme 2, (S5c) cavitation scheme 3, and (S5d) cavitation scheme 4.

Confusion Matrices for the kNN Model
are used in the table.
scheme 2. The data from FigureS12band S13b are used in the table.

Figure S14 .
Figure S14.Box-and-Whisker plots of bootstrap sample means for each model (i.e.kNN and ECOC SVM) and cavitation scheme.

Table S2 .
Number of training and testing points, as a function of temperature and cavitation level, for the kNN Model using cavitation scheme 1.The k = 3 (worst performing value of k) data from Figures S3a and S8a are used in the table.

Table S3 .
Number of training and testing points, as a function of temperature and cavitation level, for the kNN Model using cavitation scheme 1.The k = 5 data from Figures S4a and S9a are used in the table.

Table S4 .
Number of training and testing points, as a function of temperature and cavitation level, for the kNN Model using cavitation scheme 2. The k = 5 data from Figures S4b and S9b are used in the table.