MIITRA atlas: high resolution FOD template of the older adult brain

Background High quality fiber orientation distribution (FOD) template of the older adult brain facilitates white matter studies in older adults. This work compared several methods to construct an FOD template of the older adult brain in the space of the MIITRA atlas(Wu et al.,2022). Method T1w and DWI data from 400 non‐demented older adults (50% male;64.9‐98.9 yrs;54% white,43% black)(Bennett et al.,2018;Barnes et al.,2012) participating in the construction of the MIITRA atlas were used. Method 1: FOD‐based single modality registration(cost function:mean squared difference)(MRtrix3). Method 2: The nonlinear transformations generated from MIITRA T1w and DTI template construction were applied to the FODs(Wu et al.,2021). Method 3: Multichannel registration based on T1w, T2w, FA and MD information with equal weights(cost function:cross‐correlation)(ANTs). Method 4: Multichannel registration based on T1w and FOD information with equal weights(cost function:cross‐correlation)(MRtrix3). Method 5: Multichannel registration based on T1w, FA and FOD information with weights of 4,3,3 respectively(cost function:cross‐correlation)(MRtrix3). Method 6: Same as Method 5 followed by FOD registration performed on the FOD data that were already transformed by Method 5(cost function:cross‐correlation)(MRtrix3). The FOD templates generated with different methods were compared by visual inspection, image sharpness of the total apparent fiber density (AFD) map, precision of spatial normalization of FOD data used in template construction, and spatial matching with the existing MIITRA T1w and DTI templates. Result The FOD templates generated with Methods 2, 3 and 6 exhibited higher image sharpness(Fig.1A). Fiber crossings were more clearly defined in the FOD templates generated with Methods 1,4,5,6(Fig.1B). Method 6 allowed higher ACC and lower MSD of spatially normalized FOD data, indicating higher precision in spatial normalization(Fig.2A,B). Method 2 provided the highest spatial matching with the existing MIITRA templates, followed closely by Methods 4,5,6(Fig.2C,D). Overall, the FOD template generated with Method 6 achieved high image quality, was constructed based on precisely matched FOD data, and exhibited high spatial matching to the existing MIITRA templates(Fig.3). Conclusion An FOD template of the older adult brain was constructed for the MIITRA atlas. The new template exhibits high image quality, is constructed based on precisely matched FOD data, and exhibits high sp