A modelling approach to estimate the FDG arterial plasma time activity curve in Patlak-based therapy response analysis

Objectives To model the FDG arterial plasma time activity curve (APTAC) in cancer patients, to compare it with the image derived input function (IDIF) and to verify this model in Patlak glucose metabolic rate (MR_glc) measurement and therapy response monitoring (ΔMR_glc).

Methods APTAC’s based on arterial samples in 80 patients were fitted after normalisation for FDG administered activity (AA) and initial distribution volume (iDV). The medians of these parameters were used for the population-based model. In 40 other patients (20 baseline and 20 follow-up scans) this model was validated. The population-based curve, individually calibrated by AA and iDV (APTAC_AA/iDV) or one late arterial sample (APTAC_1sample) and the individual IDIF (APTAC_IDIF) were compared to the gold standard (APTAC_sampled) using the area under the curve (AUC). The effect of the three models on lesion Patlak MR_glc and therapy effect (ΔMR_glc) estimation was assessed using APTAC_sampled.

Results Correlations between AUC of the individually calibrated population based APTAC_AA/iDV and APTAC_1sample and the individual APTAC_IDIF with the gold standard (APTAC_sampled) were 0.880, 0.994 and 0.856 respectively. For MR_glc these correlations were 0.964, 0.993 and 0.966 respectively. For therapy effect these correlations were 0.943, 0.982 and 0.927 respectively. All improvements by one late arterial sample were significant (p<0.001).

Conclusions The modelled APTAC_AA/iDV had similar performance in MR_glc and ΔMR_glc estimation as an individual APTAC_IDIF. Using one late arterial sample, this improved significantly. The proposed population-based model can be used where an IDIF is not available or serial sampling is not feasible.