MO‐G‐BRF‐09: Local and Global Function Probability of the Liver

Abstract

Purpose:To develop a local‐and‐global liver function probability model to aid in physiologically adaptive RT in patients with intrahepatic cancer.Methods:Twenty‐two patients enrolled in an IRB‐approved study had liver DCE‐MRI and Indocyanine green (ICG) clearance tests prior to, during and 1 month post RT. The percentage of ICG retention in the blood plasma 15 minutes after administration (ICG‐R15) was used as an overall liver function measure. Portal venous perfusion images were derived from DCE‐MRI. Assuming parallel architecture in the liver and considering both volume and function effects of tissue subunits, a global liver function probability Pl was assumed to be a volume‐weighted summation of local function probabilities p of subunits, excluding the subunits with p<γ, a threshold indicting poor function. The local function probability p was mapped from portal venous perfusion F by a logistic function p(F)=1/(1+(F50/F)^n), where we considered function probabilities saturated at high perfusion and suppressed at low perfusion. Finally, Pl was linearly fitted to a logarithm of ICG‐R15 to determine parameters (F50, n, γ). The leave‐m‐out technique was used to cross‐validate the model. This model was applied to evaluate the doseresponse of the local function probability one month after RT.Results:The best fitted values of (F50, n, γ) were 65.3ml/(100g.min), 2.86 and 0.14, respectively. The leave‐one‐out and 10‐fold tests resulted in 66.34±3.21 and 69.26±}10.35ml/(100g.min) of F50, 2.90±0.76 and 3.25±1.59 of n, and 0.14±0.03 and 0.13±}0.05 of γ, respectively. The derived Pl significantly correlated with ICG‐R15 logarithm (r=0.66, P<;1e‐6). Average percentage reduction in the local function probability post‐RT was related to the local dose by a logistic function (P=1e−6).Conclusion:The local‐and‐global liver function probability model has the potential to guide dose redistribution during a course of treatment to spare liver functional subunits, thereby allowing target d5 ose intensification without increasing the risk for complication.The work was supported by NIH/NCI grant RO1CA132834.