TABLE 2

Sixteen Parameterizations of Logistic Growth Model of Aβ Accumulation Used to Analyze Chronological 18F-AV-45 SUVr PET Data at Regional Level

ModelK (SUVr)r (y−1)T50 (y)NSParametersSSQΔBICi
1GlobalGlobalGlobalGlobal43,073.781,500
2GlobalLocalGlobalGlobal932,273.761,600
3LocalGlobalGlobalGlobal931,324.024,200
4GlobalGlobalLocalGlobal931,245.719,900
5GlobalGlobalGlobalLocal931,147.214,200
6LocalLocalGlobalGlobal1821,131.414,300
7LocalGlobalLocalGlobal1821,079.311,000
8GlobalLocalLocalGlobal1821,070.210,400
9GlobalGlobalLocalLocal1821,002.65,910
10GlobalLocalGlobalLocal182977.04,120
11LocalGlobalGlobalLocal182920.60
12LocalLocalLocalGlobal2711,046.99,890
13LocalGlobalLocalLocal271918.9865
14GlobalLocalLocalLocal271918.8861
15LocalLocalGlobalLocal271911.0267
16LocalLocalLocalLocal360908.71,090

  • SSQ = sum of squared residuals; ΔBICi = difference in BIC between model 11 and all other models.

  • Ninety cortical and subcortical regions were included, and parameters were either restricted to single value across all regions (global) or fitted individually for each region (local). ΔBIC gives measure of parsimony of each model in relation to smallest BIC value. Model 11 (local K, global r, global T50 and local NS) gives most parsimonious fit to data.