Page 143 - GTM-4-1

Global Translational Medicine MRgFUS sonication parameters prediction

Appendix

The convolutional neural network code snippet to estimate optimal magnetic resonance imaging-guided focused
ultrasound temperature.
# Function to build the neural network model

build_model <- function() [
model <- keras_model_sequential()
model %>%
layer_dense(units = 64,
input_shape = dim(x_train)(2),
kernel_regularizer = regularizer_l2(l = 0,001)) %>%
layer_activation_relu() %>%
layer_dense(units = 128, activation = ‘relu’) %>%
layer_dropout(0.6) %>%

layer_dense(units = 64, activation = ‘relu’) %>%
layer_dense(units = 1)
model %>% compile(
loss = “mse”,
optimizer = optimizer_rmsprop(),
metrics = list(“mean_absolute_error”)
model

model <- build_model()
print_dot_callback <- callback_lambda(
on_epoch_end = function(epoch, logs) [
if (epoch %% 80 == 0) cat(“\n”)
cat(“.”)
early_stop <- callback_early_stopping(monitor = “val_loss”, patience = 20)
epochs <- 200
# Model training
history <- model %>% fit(

x_train,
y_train,
epochs = epochs,
validation_split = 0,2,
verbose = 1,
callbacks = list(early_stop, print_dot_callback)

Volume 4 Issue 1 (2025) 135 doi: 10.36922/gtm.5419