HIGH-THROUGHPUT RAT BRAIN PET IMAGING AND AUTOMATIC SPATIAL NORMALIZATION OF THE DOPAMINE D2/3 RECEPTOR LIGAND [18F]FALLYPRIDE

Objectives: Rat Brain Positron Emission Tomography (PET) imaging is often quite labor- and time intensive, as only single animals are scanned at a time and isotopic decay offers a limited window of optimal scan time. To optimize the throughput, we created a 2x2 rat holder into the human head High-Resolution Research Tomography (HRRT) scanner, enabling the scanning of four animals at a time. This higher throughput shifts the bottle-neck towards the analysis of the PET images. There is an unmet need in preclinical brain PET analysis to create reliable automated methods of spatial normalization, because manual alignment and normalization is time-consuming and inevitably operator biased. We present a non-biased standardized method for automatic spatial normalization of multimodal (CT, MR and PET) scans for the radioligand [18F]fallypride. Similar approaches are already done ([18F]FDG) or are planned ([18F]MHMZ). Methods: [18F]Fallypride was synthesized using standard procedures and obtained a molar radioactivity of over 40 GBq/µmol. Rats were anesthetized with isoflurane, placed in a customized 2x2 holder, injected with [18F]fallypride and scanned in the HRRT scanner for 45 min post injection. Up to 12 rats (three times four) were scanned with the same tracer production. Using brain PET, and a standard MR and CT image, an image template in standard space was created. In combination with this template, we created an automatic spatial normalization and VOI extraction algorithm based on MATLAB, FSL and PMOD. The non-displaceable binding potential (BPND) was calculated using a delayed scan logan plot (Tantawy et al. 2009). The automated algorithm was further assessed by transforming the PET template back to the original image, calculating the mean voxel displacement. Lastly, we used the holder and automatic procedure to measure drug induced occupancy at the dopamine D2 receptor. Values are reported ± standard deviation. Results: Nine [18F]fallypride baseline PET Scans were used to generate the PET template, which were the basis for the automatic procedure for spatial normalization. The BPND of 18 [18F]fallypride baseline scans were then compared between manually and automated spatial normalization. By using the same VOI-template the correlation between the automated and manual analyzed BPNDs in the ventral and dorsal striatum as well as mPFC (Medial Prefrontal Cortex) was R2=0.8 (ventral striatum: 2.40±0.44 automatic and 3.33 ± 1.39 manual, dorsal striatum 3.70±0.80 automatic and 5.23±2.41 manual, mPFC: 1.01±0.23 automatic and 0.83±0.64 manual). The back transformation gave a mean voxel displacement of -0.44±0.83 mm, however three out of 18 transformations failed. Conclusions: The automated analysis underestimated the BPND compared with the manual analysis, however the manually analyzed scans have a higher standard deviation, suggesting some degree of operator bias. The voxel-displacement succeeded since it is far lower than the voxel size (1.21875 mm3). In summary, it can be stated that we generated a fast and reliable procedure for reproducible spatial normalization of rat PET images for [18F]fallypride. The method has a high potential for being applicable to images from other radioligands with sufficient spatial information in the future like [18F]MHMZ