Developing a PET normal brain template using diffusion tensor imaging images: A proof of concept

Abstract

Background: Registered Positron emission tomography (PET) brain images to the standard normal PET brain templates can be performed to diagnosis dementia by using a vendor software, in which the brain template is based on T1-Weighted (T1W) images. However, the imperfection of an overlap between PET images and the PET-T1W based brain template could be observed. Objectives: This pilot study aimed to develop a new PET brain template and compare the accuracy of image registration between a conventional PET-T1W based brain template and our proposed PET-DTI based brain template. Materials and methods: The new PET-DTI based brain template was developed from twenty-four normal volunteers (age ranged 42-79 years old) who underwent 11C-Pittsburgh compound B PET scans and both T1W and diffusion tensor image (DTI) magnetic resonance imaging brain scans. The correction of Eddy-Current distortions and related artifact removing in DTI images were performed using the open-source FMRIB Software Library (FSL) to generate whole-brain probabilistic tractography maps (MRI-Probtract). MRI-Probtract map was then deformably registered and normalized to PET images, which were used for brain boundary guidance. The accuracy of image registration was assessed by applying the newly developed PET-DTI brain template to PET images of four mild cognitive impairment patients who underwent the same brain-scanning protocols. The accuracy of image registrations using the conventional PET-T1 and PET-DTI templates was evaluated qualitatively by three nuclear medicine physicians. Wilcoxon Signed Ranks test was used to compare registration scores of the two methods. Additionally, the dice similarly coefficient was obtained to quantitatively evaluate the accuracy of image registration. Results: The registration scores of the PET images registered with the PET-DTI template were significantly higher than the PET-T1 template at p-value < 0.05. This result is consistent with the dice similarly coefficient where the value of PET-DTI template was higher. Conclusion: Result of this pilot study showed that new PET-DTI brain template provides higher registration quality, suggesting the feasibility of using PET-DTI template in a clinical PET study of the brain.