ABSTRACT Introduction FDOPA PET has been used extensively to image the human brain in many clinical disorders and has the potential to be used for patient stratification and individualized treatment. However, to reach its full and effective clinical translation, FDOPA PET requires both a robust data infrastructure and analytical protocol that are capable of ensuring high quality data and metadata, accurate biological quantification, and replicable results. In this study we evaluate a digital data repository and automated analytical framework for FDOPA PET neuroimaging that can produce an individualised quantification of dopamine synthesis capacity in the living human brain. Methods The imaging platform XNAT was used to store the King’s College London institutional brain FDOPA PET imaging archive, alongside individual demographics and clinical information. A fully automated analysis pipeline for imaging processing and data quantification was developed in Python and integrated in XNAT using the Docker technology. Reproducibility was assessed in test-retest datasets both in controls and patients with psychosis. The agreement between the automated analysis estimates and the results derived by the manual analysis were compared. Finally, using a sample of healthy controls (N=115), a sensitivity analysis was performed to explore the impact of experimental and demographic variables on the FDOPA PET measures. Results The final data repository includes 892 FDOPA PET scans organized from 23 different studies, collected at five different imaging sites. After removing commercials studies, the infrastructure consisted of 792 FDOPA PET scans from 666 individuals (female 33.9%, healthy controls 29.1%) collected from four different imaging sites between 2004-2021. The automated analysis pipeline provided results that were in agreement with the results from the manual analysis, with a Pearson’s correlation that ranged from 0.64 to 0.99 for Ki cer , and from 0.79 to 1.00 for SUVR. The mean absolute difference between the two pipelines ranges from 3.4% to 9.4% for Ki cer , and from 2.5% to 12.4% for SUVR. Moreover, we found good reproducibility of the data analysis by the automated pipeline (in the whole striatum for the Ki cer : ICC for the controls = 0.71, ICC for the psychotic patients = 0.88). From the demographic and experimental variables assessed, gender was found to most influence striatal dopamine synthesis capacity ( F = 10.7, p &lt;0.001 ), with women showing greater dopamine synthesis capacity than men, while the effects of weight, age, injected radioactivity, and scanner, varied by brain region and parameter of interest. Conclusions Combining information from different neuroimaging studies has allowed us to test comprehensively the automated pipeline for quantification of dopamine synthesis capacity using FDOPA PET data and to validate its replicability and reproducibility performances on a large sample size. This validation process is a necessary methodological step for the development of the clinical application of FDOPA PET as precision medicine biomarker. The proposed infrastructure is generalisable behind the FDOPA radiotracer.

Digital data repository and automatic analysis framework for FDOPA PET neuroimaging