MR‐template‐based volumes‐of‐interest generation in individual brain PET of mouse Alzheimer model not only without MR but also without spatial normalization

Background The spatial normalization (SN) of individual mouse brain PET or MRI onto template space is one prerequisites for precise quantitative analysis positron emission tomography (PET) brains. This has recently been exploited by deep learning-based approaches, but still challenging. Moreover, most them are based on SN using MR templates, which not always available in both clinical and preclinical studies. Method In this study, we propose a new only PET-based image generation approach to resolve MR-less problem. We generated individual-brain-specific cortical volumes-of-interest (VOI) inverse-spatial-normalization (iSN) convolutional neural network (deep CNN) models with input images instead images. applied the proposed methods mutated amyloid precursor protein presenilin-1 model Alzheimer’s disease. Eighteen mice underwent T2-weighted F-18 FDG scans twice, before after administration human immunoglobulin antibody-based treatments. Specifically, skull-stripped were used as inputs iSN-based (target) cerebellar (reference) VOIs labels. compared our CNN-based VOI conventional (template-based) qualitative (visual) assessment correlation mean SUVR (i.e., uptake normalized uptake) estimated methods. Result visual assessment, there was no visible difference between template-based VOI. count obtained highly significantly (P<0.001) correlated target cortex). Conclusion Since method successfully VOI, showing concordant results that analysis, believe have established without also MR.