Uncertainty estimation in dynamic contrast-enhanced MRI.

Area of Science:

• Medical Imaging
• Biophysics
• Radiology

Background:

• Dynamic contrast-enhanced MRI (DCE-MRI) is used to estimate pharmacokinetic (PK) parameters reflecting physiological properties, particularly in tumors.
• Errors in DCE-MRI data propagate non-trivially to PK parameters, complicating interpretation.
• Quantifying these uncertainties is crucial for accurate clinical assessment.

Purpose of the Study:

• To develop and evaluate a method for calculating uncertainty maps of PK parameters derived from DCE-MRI.
• To assess the impact of various error sources on PK parameter uncertainty.
• To validate the proposed method using simulations and in vivo data.

Main Methods:

• A multivariate linear error propagation method was developed to calculate PK parameter uncertainty maps.
• The modified Kety model was used to investigate PK parameter uncertainties.
• Monte Carlo simulations and in vivo brain tumor data were employed for evaluation.

Main Results:

• The proposed method accurately estimated PK parameter uncertainties arising from noise in dynamic data.
• Uncertainties due to up to 15% signal/T1 map noise and arterial input function amplitude errors were well-estimated.
• The method showed less accuracy for errors in bolus arrival time, with significant disagreements for K(trans), ve, and vp.

Conclusions:

• The developed method offers an efficient way to calculate uncertainty maps for PK parameters in DCE-MRI.
• The method is applicable to various sources of uncertainty, enhancing the robustness of DCE-MRI analysis.
• Accurate uncertainty quantification is vital for reliable interpretation of DCE-MRI derived parameters in clinical settings.