Direct parametric reconstruction from undersampled (k, t)-space data in dynamic contrast enhanced MRI.

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) provides insights into tissue microvasculature and rapid concentration changes post-contrast agent injection.
• Traditional indirect methods reconstruct individual DCE images and fit pharmacokinetic models, which can be susceptible to image degradation with undersampling.
• Accelerated acquisition techniques like kt-FOCUSS are used to mitigate undersampling artifacts in conventional DCE-MRI reconstruction.

Purpose of the Study:

• To develop and evaluate a Bayesian inference framework for direct estimation of functional parameters from undersampled (k, t)-space DCE-MRI data.
• To compare the accuracy and diagnostic utility of directly estimated kinetic parameters against indirectly derived parameters.

Main Methods:

• A Bayesian inference framework was implemented to directly estimate pharmacokinetic parameters using the extended Tofts model from undersampled (k, t)-space DCE-MRI data.
• The proposed method was evaluated on simulated abdominal DCE phantom data and real prostate DCE data, including fully sampled and 4- and 8-fold undersampled datasets.
• A peripheral zone prostate cancer diagnostic model was used to assess the impact of the methods on cancer diagnosis probability mapping.

Main Results:

• Directly estimated kinetic parameters showed significantly better correspondence to ground truth parameters in simulated data, with up to 70% higher mutual information compared to indirect methods.
• For prostate DCE data, the direct kinetic parameters provided a clearer depiction of tumor morphology.
• The evaluation using a prostate cancer diagnostic model indicated potential improvements in diagnostic accuracy with the direct estimation method.

Conclusions:

• The proposed Bayesian inference framework offers a direct and accurate method for estimating kinetic parameters in DCE-MRI, particularly beneficial for undersampled acquisitions.
• Direct kinetic parameter estimation enhances the depiction of tissue microvasculature and tumor morphology, potentially improving diagnostic performance in DCE-MRI.
• This approach represents a significant advancement in DCE-MRI analysis, enabling more precise functional parameter estimation and better characterization of disease.