Quantifying neocortical structural changes for clinical trials in Alzheimer’s disease: Comparison between tensor-based morphometry and longitudinal freesurfer


      Accurate and reliable quantification of structural brain changes, within a regulatory compliant framework, is important in clinical trials for Alzheimer’s disease (AD). Cortical changes relevant to AD pathology have been reported in various regions including isthmus cingulate, precuneus, inferior parietal, temporal pole and prefrontal cortex[1,2]. We report the performance of atrophy measurements for these structures using tensor-based morphometry (TBM)[3], and compare to changes in cortical thickness measured with longitudinal FreeSurfer (LFS_Th)[4] on the standardized ADNI dataset[5].


      3DT1 MRI sequences from ADNI-1/2 were analyzed using pair-wise approaches to assess changes in volume (TBM) and cortical thickness (LFS_Th). Thickness changes over volume from the LFS processing suite was chosen for comparison due to lower variability (data not presented). Baseline and Month-3 data from 20 ADNI-2 normal controls (NCs) were used for test-retest purposes. Absolute symmetrized percent change (ASPC)[6] was calculated to assess variability. Baseline, Month-12 and Month-24 ADNI-1 data were used to quantify percent changes for 493 subjects (99 ADs, 115 LMCI-converters, 115 LMCI-non-converters and 164 NCs). Generalized areas under curve (AUCs) were calculated and the DeLong test[7] was performed to compare receiver operating characteristic (ROC) curves between methods. Linear regression was performed to test bias.


      Mean ASPC (%) between Baseline_Month-3 scans from NCs were smaller for TBM than LFS_Th for all regions, indicating lower variability with the TBM method (33% for inferior parietal up to 56% fortemporal pole). Generalized AUCs and ROC results for Month-12 and Month-24 changes are reported in Tables 1 and 2. The TBM method yielded systematically larger AUCs and better ROC discrimination between several category pairs. Linear regression of Month-12 and Month-24 changes yielded intercepts close to zero for all regions and subgroups, indicating negligible bias for either TBM or LFS_Th methods.


      The TBM method successfully and reliably quantified changes over time for various cortical regions, and showed improved sensitivity to differentiate subgroups compared to LFS_Th. A pair-wise analysis approach to quantify changes using TBM may be a useful option for clinical trial applications, considering the intrinsic limitation of LFS_Th which needs to be run after scans from all required visits are acquired. References: [1] McEvoy, et al. Radiology. 2009;251(1):195-205. [2] Greene, et al. Neurobiol Aging. 2010;31(8):1304-1311. [3] Vercauteren, et al. Med Image Comput Comput Assist Interv. 2008;11:754-761. [4] Fischl, et al. Cerebral Cortex. 2004;14:11-22. [5] Wyman, et al. Alzheimers Dement. 2013;9(3):332-337. [6] Reuter, et al. Neuroimage. 2012;61(4):1402-1418. [7] DeLong, et al. Biometrics. 1988;44(3):837-845.
      Table 1Generalized AUCs for each cortical region and method at M12 and M24
      TBM Month-12LFS Th Month-12TBM Month-24LFS Th Month-24
      Temporal pole0.720.640.760.69
      Isthmus cingulate0.690.610.740.63
      Interior parietal0.640.610.660.63
      Note: Generalized AUCs were calculated by computing probability that two random subjects are properly ranked with respect to ordinal outcome with two or more levels.
      Table 2DeLong test results (AUC-TBM/AUC-LFS_Th (P value)) in comparing groups at M12 and M24
      AD/NC Month-12AD/NC Month-24AD/LMCI-c Month-12AD/LMCIc Month-24LMCI-c/LMCI-nc Month-12LMCI-c/LMCI-nc Month-24
      Temporal pole0.84/0.76 (0.013)0.92/0.82 (0.002)nsns0.73/0.57 (<0.001)0.73/0.64 (0.014)
      Isthmus cingulate0.81/0.67 (0.001)0.88/0.72 (< 0.001)0.59/0.51 (0.041)nsns0.72/0.60 (0.002)
      Precuneus0.77/0.68 (0.007)0.82/0.70 (< 0.001)nsnsns0.69/0.61 (0.010)
      Interior parietal0.73/0.67 (0.054)0.79/0.73 (0.060)nsnsnsns
      Prefrontal0.70/0.60 (0.001)0.82/0.77 (0.065)0.57/0.54 (0.041)nsnsns
      Note: ns (Not Significant), LMCI-c (Late MCI-converter), LMCI-nc (Late MCI-non-converter).