Table 2. The classification accuracy, sensitivity, and specificity of the support vector machine (SVM) classifiers are given for distinguishing normal controls (NC) and subjects with Alzheimer's disease (AD), NC and subjects with mild cognitive impairment (MCI), and subjects with MCI and AD.
NC.vs. AD | NC.vs. MCI | MCI.vs. AD | |
Hippocampal Volumes and Shapes | |||
Hp volumes | 65.5% (CI: 64.4%∼66.6%) Sensitivity = 57.8%, Specificity = 73.3% | 61.9% (CI: 61%∼62.8%) Sensitivity = 57.7%, Specificity = 66.1% | 42.3% (CI:32.2%∼52.3%) Sensitivity = 45.3%, Specificity = 39.2% |
Hp shapes | 79.2% (CI: 78.5%∼80%) Sensitivity = 75.8%, Specificity = 82.8% | 67.4% (CI: 66.2%∼68.6%) Sensitivity = 64%, Specificity = 70.8% | 57.2% (CI: 51.7%∼62.8%) Sensitivity = 59.9%, Specificity = 54.5% |
Lateral Ventricular Volumes and Shapes | |||
LV volumes | 65.5% (CI: 64.9%∼66.1%) Sensitivity = 58.8% Specificity = 72.3% | 63.1% (CI: 62%∼64.2%) Sensitivity = 50.9%, Specificity = 75.3% | 30.4% (CI: 18.8%∼42%) Sensitivity = 21.2% Specificity = 39.7% |
LV shapes | 61.5% (CI: 59.2%∼63.9%) Sensitivity = 59.2%, Specificity = 63.9% | 59% (CI: 56.1%∼62%) Sensitivity = 53.6%, Specificity = 64.4% | 60.1% (CI:57.1%∼63.1%) Sensitivity = 62%, Specificity = 58.3% |
CSF | |||
CSF markers | 81.4% (CI: 80.3%∼82.5%) Sensitivity = 87.4%, Specificity = 74.9% | 68.4% (CI: 67.8%∼69%) Sensitivity = 66.7%, Specificity = 70.1% | 61.3% (CI: 59.4%∼63.2%) Sensitivity = 84.3%, specificity = 38.3% |
Combination | |||
CSF, Hp volumes, LV volumes | 85.4% (CI: 84.3%∼86.5%) Sensitivity = 88.8%, Specificity = 82% | 72% (CI: 70.5%∼73.5%) Sensitivity = 70.1%, Specificity = 73.9% | 60.9% (CI: 59.1%∼62.8%) Sensitivity = 80.4%, Specificity = 41.4% |
CSF, Hp shapes, LV shapes | 92.2% (CI: 91%∼93.5%) Sensitivity = 94.7%, Specificity = 89.8% | 70.3% (CI: 68.6%∼72.9%) Sensitivity = 69.5%, specificity = 71.9% | 69.6% (CI: 66.4%∼72.8%) Sensitivity = 70.7%, Specificity = 68.6% |
The volumes and shapes of the hippocampus (Hp) and lateral ventricles (LV) as well as cerebral spinal fluid (CSF) markers are respectively used as features in the SVM.