Table 3

Prediction performance for ADAS-13 scores*

ModelHippocampal inputCortical thickness inputCombined hippocampal and cortical thickness input
rRMSErRMSErRMSE
ADNI1
 Linear regression with lasso0.22 ± 0.118.72 ± 0.810.56 ± 0.087.44 ± 0.720.56 ± 0.087.42 ± 0.74
 Support vector regression0.23 ± 0.118.70 ± 0.850.52 ± 0.087.68 ± 0.760.53 ± 0.087.62 ± 0.78
 Random forest regression0.15 ± 0.109.27 ± 0.800.54 ± 0.087.55 ± 0.760.54 ± 0.087.51 ± 0.77
 APANN0.53 ± 0.097.56 ± 0.760.51 ± 0.107.67 ± 0.760.60 ± 0.087.11 ± 0.72
ADNI2
 Linear regression with lasso0.14 ± 0.119.69 ± 0.700.61 ± 0.077.77 ± 0.710.61 ± 0.077.78 ± 0.71
 Support vector regression0.21 ± 0.109.75 ± 0.790.63 ± 0.077.65 ± 0.680.63 ± 0.077.66 ± 0.70
 Random forest regression0.24 ± 0.099.77 ± 0.760.58 ± 0.077.97 ± 0.650.58 ± 0.087.97 ± 0.67
 APANN0.52 ± 0.078.32 ± 0.790.63 ± 0.077.58 ± 0.710.68 ± 0.067.17 ± 0.71
ADNI1 + 2
 Linear regression with lasso0.12 ± 0.089.37 ± 0.500.58 ± 0.067.71 ± 0.480.58 ± 0.067.71 ± 0.48
 Support vector regression0.18 ± 0.079.39 ± 0.540.59 ± 0.057.65 ± 0.420.59 ± 0.057.65 ± 0.42
 Random forest regression0.18 ± 0.099.63 ± 0.610.57 ± 0.057.76 ± 0.460.57 ± 0.057.75 ± 0.46
 APANN0.54 ± 0.067.99 ± 0.590.57 ± 0.057.79 ± 0.510.63 ± 0.057.32 ± 0.53

  • ADNI = Alzheimer’s Disease Neuroimaging Initiative; APANN = anatomically partitioned artificial neural network; RMSE = root mean squared error; SD = standard deviation.

  • ↵* Findings are presented as mean ± SD.