Abstract
1461
Purpose: This study was performed to develop a novel quantification method for shape features on amyloid PET and brain MRI and evaluate its prognostic significance in MCI patients from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort.
Methods: We assessed 219 MCI patients (M/F=118/101, mean age, 71.4 ± 7.1 years) from ADNI database. AV45_BASS (brain amyloid smoothing score) was defined as inverse spherocity corrected surface/volume ratio which segmented with SUV threshold of 50 % in AV45 PET image. MR_BAI (brain atrophic index) was defined as spherocity corrected surface/volume ratio in MR image. Shape features multiplied AV45_BASS and MR_BAI. Parameters including AV45_BASS, MR_BAI, and shape features were analyzed for AD conversion with clinical parameters and cognitive measurements. Univariate and multivariate analyses for AD conversion were performed.
Results: Ninty-four (42.9%) of 219 patients were converted to AD during follow-up period for 48 months. Age, MMSE, ADAS-cog, APOE, SUVR, AV45_BASS, MRI_BAI, shape features were significantly higher in patient with AD conversion (P<0.01). Shape features were strongly correlated with SUVR. Results of univariate analysis showed that age, MMSE, ADAS-cog, APOE, SUVR, AV45_BASS, MR_BAI, shape features correlated with the rate of AD conversion (P<0.01). In the multivariate analyses, high shape features and ADAS-cog were associated with an increased risk of AD conversion (P=<0.0001, <0.0001, respectively, hazard ratio=5.58, 4.53 respectively). Shape features were strongly correlated with the longitudinal change in cognitive measurements.
Conclusions: New quantification method of shape features on amyloid PET and brain MRI was developed. Shape features on amyloid PET and brain MRI was the strongest prognostic factor for predicting AD conversion in MCI patients from ADNI cohort.
Multivariate Cox regression analysis of potential MCI conversion to AD factors influencing MCI perio
