THE MEASUREMENT OF HIPPOCAMPAL ATROPHY RATES WITH MRI FOR A 3-YEAR STUDY APPEARS TO BE AT LEAST 3 TIMES MORE SENSITIVE THAN A 1-YEAR STUDY BASED ON BACK-TO-BACK REPRODUCIBILITY

Hippocampal atrophy rate measurements from MRI scans are often used as a biomarker of Alzheimer's disease including in clinical trials. This study rigorously assessed how much the signal to noise of hippocampal atrophy rates is improved by comparing the measurements over intervals of 1 and 3 years. While rarely mentioned in the literature, as part of the first Alzheimer's Disease Neuroimaging Initiative (ADNI1), two 3D T1 weighted MPRAGEs were acquired back-to-back (BTB) during each subject visit - with the acquisition of the second MPRAGE usually starting within seconds of completion of the first - providing an excellent data set for reproducibility and noise measurements. To measure the noise of the hippocampal atrophy rate, the reproducibilities of 5 fully automated methods - FSL/FIRST 5.0.4, FreeSurfer 5.3.0 cross sectional, FreeSurfer 5.3.0 longitudinal, Multiple-Atlas Propagation and Segmentation algorithm (MAPS) independently and MAPS combined with the boundary shift integral (MAPS-HBSI) - were calculated. The 264 patients in the ADNI1 study with BTB 1.5T MPRAGE MRIs at baseline, year 1 and year 3 were used. The BTB reproducibility of the atrophy rate of both the left and right hippocampus over 1 year and 3 years were compared with a simple statistical test based on the binomial sign test that handled outlying data points robustly. The noise in the atrophy rate, as measured by the reproducibility in units of percentage points, showed no statistical difference between 1 and 3 years in 4 of the 5 methods. FSL/FIRST was slightly noisier in both hippocampi over 1 year than 3 years (p=0.002 & p=0.002). For example, the noise for the left hippocampus for FreeSurfer longitudinal was 2.0% over 1 year and 2.2% over 3 years while the corresponding atrophy rates were 1.9% and 5.2%. The noise of measuring hippocampal atrophy rates over 3 years was no worse than over 1 year for the 5 methods studied. Therefore, if the atrophy rate was stable over the 3 years, the 3 year study would be at least 3 times as sensitive as the 1 year study to changes in atrophy rate.