FIGURE 3.
11 ELISA parameters (11 E) and strain data from mechanical tests (in the form of invariants I 1, I 2,… ) form the input of an extended type of constitutive artificial neural network. Machine learning adjusts its stress output to the one measured experimentally. Thereby, the neural network learns to describe the mechanical behavior of brain tissue and to predict it from the ELISA parameters. The extended CANN consists of a standard CANN block and a parallel deep neural network computing the Prony parameters of the viscoelastic constitutive behavior. Note that the stress response P is computed by a recursive update in time t. The hidden neurons in the network are illustrated as blank circles, while the associated network weights are depicted as black arrows.