Sang Ho Lee PhD, Presenter: Nothing to Disclose

Yasuji Ryu MD, Abstract Co-Author: Nothing to Disclose

Koichi Hayano MD, Abstract Co-Author: Nothing to Disclose

Satoshi Kobayashi MD, Abstract Co-Author: Nothing to Disclose

Kazuto Kozaka MD, Abstract Co-Author: Nothing to Disclose

Hiroyuki Yoshida PhD, Abstract Co-Author: Patent holder, Hologic, Inc Patent holder, MEDIAN Technologies

The dual-input FL-PK model using CAP is potentially feasible mitigating the uncertainty of parametric-fitting in liver 4-phase DCE-CT.

DCE-CT imaging protocol is a trade-off between the number of data points and total radiation dose. Considering availability and radiation exposure, 4-phase DCE-CT is a favored option in clinical practice for liver imaging. However, the 4-phase DCE-CT has difficulty in modeling pharmacokinetics due to the low temporal resolution. The aim of this study was to investigate the feasibility of dual-input flow-limited pharmacokinetic (FL-PK) modeling using convolution area property (CAP) in liver 4-phase DCE-CT.

We replaced the infinite integral in CAP with a finite integral on the interval defined by the sampling time from the arterial to delayed phases, so that the CAP was valid on a finite imaging time period. This time-variant scheme is to modify the fundamental assumption that the capillary-tissue system is modeled as a linear time-invariant (LTI) system. However, a dual-input FL-PK model with the LTI system unstably fitted the data because of high uncertainty in the intervals of data points that might contain flow information. Applying the CAP with the dual-input FL-PK model allowed realistic parameter estimates while well distinguishing the arterial and portal flow between HCC and normal liver parenchyma with a lower curve-fitting error.

We investigated the 4-phase liver DCE-CT data of nine patients with hepatocellular carcinoma (HCC). Using bolus tracking technology (Smart Prep; GE Healthcare), the arterial, portal and delayed phase scans were initiated 17s, 70s and 150s after the threshold of 200HU was attained in the aorta. The arterial and portal input curves were modeled by a sum-of-exponentials function. A dual-input FL-PK model was used to calculate liver kinetic parameters. The CAP that describes the product of area under individual functions was applied between the dual-input function and the tissue residue function of the FL-PK model, followed by its differentiation with use of the product rule. The number of curve-fitting parameters was limited to 4 for avoiding over-fitting to the data.

Lee, S, Ryu, Y, Hayano, K, Kobayashi, S, Kozaka, K, Yoshida, H, Feasibility of Dual-input Flow-limited Compartmental Pharmacokinetic Modeling Using Convolution Area Property in Liver 4-Phase DCE-CT.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14008216.html