Fig. 1

Schematic of experimental and computational methods. Panel a shows our model pre-processing work flow. Images are first registered using a rigid registration algorithm. DW- and DCE-MRI are then used to estimate \( {\phi}_T\left(\overline{x},t\right) \) and \( {\phi}_V\left(\overline{x},t\right) \) at each time point. Using data from t₁ through t₅, all 10 models are calibrated (panel b) to return estimates of the model parameters that minimize the error between the measurement and model. The AIC is then calculated and used to select the most appropriate model (panel c) that balances model complexity and model fit. The selected model is then used in a forward evaluation to provide a “forecast” (panel d) of future \( {\phi}_T\left(\overline{x},t\right) \) and \( {\phi}_V\left(\overline{x},t\right) \) at t₆ and t₇. The predicted \( {\phi}_T\left(\overline{x},t\right) \) and \( {\phi}_V\left(\overline{x},t\right) \) are then compared directly back to the measurement obtained from DW- and DCE-MRI