Optimized atlas for early tau-PET staging via native space segmentations.

Positron Emission Tomography (PET) early Braak staging might be susceptible to anatomical variability and atrophy in the medial temporal lobe (MTL) structures. These factors should be accounted for in an optimized atlas to improve staging accuracy. This study aimed to compare the accuracy of early tau detection using traditional standard space methods versus using a native space MTL segmentations. Twelve native space MTL structures were used as regions of interest (ROI) for [18F]MK6240 tau-PET images and compared with standard space Braak stage ROIs for 333 participants aged over 55. We used the Rey Auditory Verbal Learning Test (RAVLT) to assess memory function. Native and standard space tau-PET stage ROIs were compared, then combined with anatomical constraints into an optimized standard space MTL atlas. The native space MTL tau-PET staging identified 34 participants with significantly more advanced tau accumulation. Of these, 14 had significant entorhinal and transentorhinal tau despite being classified as Braak stage I when using the original standard space method (here called pre-I stage). In addition, 19 were classified as Braak stage III despite being at Braak stage II using standard space methods (here called pre-III stage). These pre-III participants displayed a significant memory impairment. We found that a standard space spatial smoothing to 6 mm at FWHM best allowed to replicate native space results, resulting in the optimized atlas identifying 29 of these 33 more advanced cases. Therefore, standard space approaches can be improved to better capture early AD tau pathology and be more sensitive to cognitive impairment.