Interoperability of the MIITRA atlas with complementary atlases: development of a comprehensive array of gray matter labels

Abstract

AbstractBackgroundIn order to enhance the functionality of the MIITRA atlas and its interoperability with complementary atlases, a comprehensive set of gyral‐based, cytoarchitecture‐based, and functional connectivity‐based gray matter labels were constructed in MIITRA space.MethodT1w images from the 400 older‐adults included in the construction of MIITRA atlas were nonlinearly registered using ANTs SyN to the MNI152 6thgen,MNI Colin, and the ICBM2009c templates, and volumetric labels from Harvard‐Oxford,Julich,AAL3,Buckner,CoBrALab,Hammersmith and Princeton Visual atlases were warped to the space of each participant. The ANTs‐derived transformations used to build the MIITRA T1w template were used to map the corresponding gray matter labels from raw space to exact physical locations in the final MIITRA space following a multi‐atlas correction scheme to correct for residual misregistration. T1w images from a separate group of 100 older‐adults for evaluation were nonlinearly normalized to the MIITRA template and the inverse transformation was applied to the gray matter labels to transform them to each individual’s space. The T1w image of each individual was nonlinearly registered to every other individual and the labels from the 99 other older‐adults were transformed to each individual. Pair‐wise Jaccard index for each label was calculated in every individual’s native space. The same procedure was followed for the labels obtained directly from the respective source templates.ResultThe overlap was higher for the AAL3 labels obtained from MIITRA with an average pair‐wise Jaccard index of 0.59 and standard deviation of 0.08, compared to 0.55±0.1 for the AAL3(Fig1A) labels obtained from the MNI Colin template. The pair‐wise Jaccard index for the Julich(Fig1C) and CoBrALab labels(Fig2A) was 0.54±0.13 and 0.75±0.17 from MIITRA and, 0.53±0.14 and 0.74±0.17 from ICBM2009c. For the Harvard‐Oxford(Fig2B), Buckner(Fig2C), Hammersmith(Fig1B) and Princeton Visual(Fig2D), the pairwise Jaccard was 0.68±0.14, 0.61±0.07, 0.71±0.10 and 0.35±0.06 respectively from MIITRA and 0.66±0.11, 0.61±0.08, 0.67±0.11 and 0.32±0.08 respectively from MNI152 template. Heat maps representing the Jaccard‐index for each template are shown in Fig3.ConclusionGray matter labels from MIITRA showed higher pair‐wise overlap suggesting that labels from MIITRA space allows more consistent segmentation of the gray matter of the older‐adult brain compared to the labels obtained from templates constructed using younger or middle‐aged data.