Modeling of a population-based input function (PBIF) using the Feng model in dynamic ⁶⁸Ga-DOTATOC whole body PET/CT scans: feasibility of shortened imaging protocols on PET/CT Vision 600 system ®.

Abstract

Background: This study focuses on modeling a population-based input function (PBIF) in dynamic ⁶⁸Ga-DOTATOC PET/CT exams, with the aim of developing clinically adoptable protocols. The PBIF is derived from an image-derived input function (IDIF), ensuring a non-invasive and standardized approach to tracer kinetic modeling.

Methods: Patients with well-differentiated neuroendocrine tumors were included from the GAPETNET clinical trial (n = 37), divided into a PBIF modeling group (n = 20) and an independent validation group (n = 17). Dynamic whole-body (dWB) PET/CT imaging was performed using a Vision 600 PET/CT system. A population-based input function (PBIF) was modeled using the Feng approach and scaled to individual patient-specific IDIFs over two different time windows (sPBIF_3 - 7: 25-55 min, sPBIF_5 - 7: 40-55 min). The scaled PBIF was normalized to IDIF data from 6 to 55 min post-injection. A full individual patient-specific IDIF using data from 0 to 70 min post-injectionwas used as the reference for AUC and Ki comparisons. IDIFs and scaled PBIFs were compared by assessing the area under the curve (AUC) and radiotracer influx rate (Ki). Linear correlation and Bland-Altman analyses were conducted for AUC and Ki comparisons. Additionally, Mann-Whitney tests were performed to compare Ki values obtained with IDIF and sPBIF in both tumoral lesions and physiological organs.

Results: The lowest mean relative AUC bias was observed with sPBIF_3 - 7, calculated to be 2.7 ± 7.9%, and was slightly higher with sPBIF_5 - 7 (7.35 ± 8.58%). The correlation coefficient (R²) with the sPBIFs was high, with a minimum of 0.95 for the sPBIF_5 - 7. When analyzing K_i metrics, biases tended to be lower with the 40-55 min time window (Mean ± SD bias = 1.61 ± 3.33 for K_{i max} and 1.64 ± 2.96 for K_{i mean}). No significant differences in K_i values was observed with the sPBIFs compared to the IDIF (p > 0.05), for either tumoral lesion or physiological organs.

Conclusion: Our study has demonstrated the feasibility the PBIF approach to estimate tumor or physiological Ki values from a shortened dWB ⁶⁸Ga-DOTATOC PET acquisition, using the Feng model.