Cardiovascular disorders are among the most common causes of death worldwide. The formation of a thrombus alters blood flow patterns and impairs blood flow. Therefore, blood flow prediction can aid in an improved understanding of various medical disorders, and developing innovative diagnostic procedures and therapies. Numerical simulations are time-consuming and computationally expensive, and this approach for complex computational domains is prohibitively costly. Therefore, the neural network method is an effective approach for the solution of the corresponding problems. This study aimed to forecast wall shear stress in arteries with varying degrees of stenosis. The prediction approach involved a multi-layer perceptron, with the hyperparameters tuned by the Slime Mould algorithm throughout the training phase. The Hemocell model was used by combining the lattice Boltzmann and immersed boundary methods to generate input datasets for the prediction model. R-squared and root-mean-squared error statistical indices were used to evaluate the outcomes of the predictions. The results indicate that the R-squared values for the testing data were 0.965, 0.959, 0.955, 0.929, and 0.903, respectively, for stenosis levels of 90%, 80%, 60%, 40%, and 20%. The article demonstrates thrombus growth relative to stenosis location across varying levels of blockage. Notably, thrombus growth predominantly occurs before the stenosis peak in lower blockage cases, while higher blockages exhibit growth primarily after the peak.
Read full abstract