Decoupling iterative algorithm for rapid calculation of physiological parameters from the standard Tofts model.

BACKGROUND

The standard Tofts model (STM) is an important pharmacokinetic model for analyzing dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) data. However, it is very time-consuming using the STM to perform pixel-by-pixel analysis for 3D DCE-MRI data.

PURPOSE: We developed a decoupling iterative algorithm, prediction-correction method (PCM), for rapid calculation of physiological parameters using the STM.

METHODS

The idea behind PCM is to eliminate the need to fit the entire contrast agent concentration (C(t)) as function of time (t) curve to calculate the volume transfer constant (Ktrans) and the volume fraction of the extravascular extracellular space (ve) using the STM. The early portion of C(t) was used to obtain Ktrans with prediced ve value, and the late portion of C(t) was used to obtain ve with predicted Ktrans. This procedure was iteratively performed until the changes of Ktrans and ve were less than the given tolerance errors. The method was first validated using the quantitative imaging biomarker alliance (QIBA) data. Then the public prostate DCE-MRI dataset that was scanned twice and the breast DCE-MRI dataset were used as applications of the PCM and compared with the conventional way of fitting the STM. The repeatability coefficients (RC) of the calculated parameters were also determined.

RESULTS

For QIBA data, there was an excellent agreement for calculating physiological parameters between the PCM and the conventional STM. For clinical data, there was a small percentage error (<10%) in the calculations of Ktrans and ve between the two methods and between two scans. Overall, the PCM was about 10 times faster than the STM for each pixel. The repeatability of calculating Ktrans and ve was similar between the PCM and STM.

CONCLUSION

The PCM significantly accelerated the calculations of Ktrans and ve with an accuracy close to the STM. By using the PCM, the physiological parameters can be calculated rapidly for 3D DCE-MRI data to aid cancer diagnosis.