- Jan 13 - Oct 18, 2019
STSM report

GFR and perfusion modelling (closed)

STSM start date: 2019/01/13
STSM end date: 2019/01/18
Grantee name: Christian Mariager

PURPOSE OF THE STSM

Perfusion and glomerular filtration rate (GFR) estimation are essential tools in the evaluation of renal function. However, in the field of MRI, clinical translation is still lacking due to various factors including consensus on model standardization and the complexity of the renal tissue itself. Therefore, direct comparisons across different studies, disease states and research institutions can be difficult to perform.

The purpose of this STSM was to learn how post processing and subsequent perfusion and GFR estimation is performed at the host institution and compare to our own protocols. This comparison is to be performed using data from healthy rats and rats subjected to unilateral ischemic reperfusion injury (IRI) across various perfusion models. Lastly, the purpose of the STSM was also to discuss standardization of perfusion and GFR measures across methods and institutions. This is done in an effort to produce a set of standard operating procedures to be used in pre-clinical perfusion and GFR modelling using 1H and hyperpolarized 13C MRI.

DESCRIPTION OF WORK CARRIED OUT DURING THE STSM

Using six sets of dynamic contrast enhanced (DCE) MRI data from a IRI rodent model as a starting point, the initial work was focused on learning the use of the software used for perfusion and GFR modelling at the host institution – the PMI software (https://sites.google.com/site/plaresmedima/downloads). The quality of the data and the extracted arterial input functions (AIFs) was assessed, see figure B). Subsequently, a customized version of PMI was created to allow for scaling of the area under the AIF in cases where the acquired data is sub-optimal, in order to facilitate future development of the post processing done at our institution using the PMI software, see figure A). After assessing additional DCE and hyperpolarized 13C urea and pyruvate data, analysis issues and possible causes for these were identified and noted for future experiments. Lastly, analysis strategies for perfusion and GFR modelling using DCE MRI and 13C hyperpolarized MRI data were discussed, and robust strategies for data acquisition and post processing for use at our home institution were proposed. However, additional work at our institution is needed in order to implement these solutions as best practices.

DESCRIPTION OF THE MAIN RESULTS OBTAINED

Upon assessment of DCE, 13C urea and pyruvate data sets using the PMI perfusion analysis software it is evident that a number of issues are present. Firstly, the aortic vessel present in the rodent data is too small to use for good AIF prediction and subsequent perfusion analysis, in part due to partial volume effects. A more robust solution could be to use a larger vessel such as the ventricle of the heart. Secondly, the typical AIF able to be extracted from the assessed data seems to be saturated, see figure B). In the case of DCE, a saturated AIF can be remedied by modifying MRI sequence parameters or by utilizing a smaller dose of gadolinium contrast agent such as 1⁄2 or 1⁄4 of the standard dose. In the case of hyperpolarized tracers, the dose cannot be reduced due to the inherent nature of the tracer and the accompanying MRI acquisition strategies. Additionally, the partial volume effects around the aortic region are particularly severe in the hyperpolarized case in part due to low resolution used. Therefore, other solutions are warranted.

A proposed solution to a workflow that can circumvent these issues is to acquire a set of perfusion data for each tracer in three healthy rats, where perfusion parameters are well known. Data should be acquired with a slice through the ventricle of the heart and through the kidneys. The slice through the ventricle is then used to produce a good “standard” AIF from which a scaling factor can be determined by integration of the AIF curve. The scaling relies on the fact that C(t) ∝ [S(t) − S0 ]/S0 , where S(t) is the signal at time t and S0 is the average signal of the baseline before the contrast agent arrives. This is referred to as absolute enhancement. The validity of the estimated perfusion parameters obtained using the AIF scaling factor is then cross referenced with the previously known perfusion parameters to ensure the accuracy of the model predictions. The produced scaling factors can then be applied to future animal studies using DCE where the aortic vessel is not sufficient to capture a good AIF, and for all perfusion assessment using hyperpolarized data, thereby alleviating the issue with partial volume effects.

An AIF scaling factor will have to be determined for each combination of MRI scanner, sequence and animal  type in order to be accurate. The work to acquire these data should be initiated in the near future. 

Lastly, an issue remains in regards to the T1 correction that must be performed prior to any perfusion assessment using hyperpolarized tracers. When the perfusion data is corrected by multiplying by a T1 dependent exponential, the washout phase of the perfusion curve tends to be distorted resulting in an errorneous model prediction. However, since we are primarily interested in filtration using tracers like 13C-urea, much of the washout phase can be exempted in fitting of the desired perfusion model.

FUTURE COLLABORATIONS

Pending the implementation of the best pratices learned during this STSM at the home institution, a future collaboration to further improve perfusion and GFR estimation using PMI and hyperpolarized tracer data is certainly possible.

Members