A Novel Cost Function in AIF Model Fitting for DCE-MRI Studies.

Poster Number: PO-GeP-M-25

Abstract ID: 53163

A Novel Cost Function in AIF Model Fitting for DCE-MRI Studies

R He*, K Wahid, B McDonald, Y Ding, A Mohamed, B Elgohari, K Hutcheson, C Fuller, S Lai, UT MD Anderson Cancer Center, Houston, TX

Multi-Disciplinary General ePoster

Category: Scientific:Multi-Disciplinary:MRI in Radiation Therapy:MRSI, DCE and Other Novel Techniques for Planning

Purpose: Current methodology in analyzing dynamic contrast enhanced (DCE)-MRI is heavily dependent on Arterial Input Function (AIF) modeling. To account for large inter- and intra-patient variabilities of AIFs, accurate and efficient modeling and utilization of patient-specific AIFs are crucial. In this abstract, a new cost function in AIF model fitting is developed that can automatically determine the arrival time (AT) and time to peak (TTP) of upslope from the data efficiently.

Methods: In this study, we implement an extension of a six-parameter linear function plus bi-exponential function AIF model. Towards the purpose of acquiring the corresponding AT and TTP time points of upslope for each AIF time series, we design a special cost function, where fitting is performed with global optimization on the AIF model function with seven parameters.

Results: Currently we have tested this new model on a dataset of over 300 images of DCE-MRI, and the model fitting was shown to be consistent and stable. In previous studies for analytic AIF modeling, AT and TTP are usually predetermined and hinder the quality of fitting results. Moreover, in many settings it is simply not feasible to perform AIF measurements reliably due to data acquisition constraints. Herein we have shown that our new method can precisely and efficiently fit the analytic models without deliberation on the determination of AT and TTP, parameters that are typically difficult to obtain accurately due to inter- and intra-patient variabilities. Additionally, since our method relies on generalized parameter estimation, it can be extrapolated to other AIF models.

Conclusion: We suggest our proposed method for modeling AIF offers a quick and reliable approach for quantitative investigation of DCE-MRI. This methodology can be adapted into current DCE-MRI workflows to improve clinical implementation of these novel imaging techniques.