Table 2.
Performance of classifiers in differentiating cognitively normal from Alzheimer’s disease participants
Model/feature set | Accuracy (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | AUC |
---|---|---|---|---|---|---|
Fine Decision Trees | ||||||
HipV | 78.3 | 80.7 | 74.7 | 82.1 | 72.9 | 0.82 |
HipV, Demographics, APOE4 | 81.7 | 85.1 | 76.9 | 83.5 | 79.1 | 0.82 |
MRIv | 82.5 | 86.0 | 77.5 | 84.6 | 79.4 | 0.81 |
MRIv, Demographics, APOE4 | 83.0 | 85.5 | 79.4 | 85.7 | 79.1 | 0.82 |
Linear SVM | ||||||
HipV | 83.4 | 88.5 | 75.9 | 84.0 | 82.1 | 0.90 |
HipV, Demographics, APOE4 | 85.2 | 88.5 | 80.3 | 86.6 | 83.0 | 0.92 |
MRIv | 76.3 | 95.8 | 48.1 | 72.3 | 88.9 | 0.94 |
MRIv, Demographics, APOE4 | 76.8 | 96.5 | 49.0 | 73.0 | 89.7 | 0.95 |
Fine KNN | ||||||
HipV | 77.8 | 78.7 | 76.3 | 82.7 | 71.4 | 0.78 |
HipV, Demographics, APOE4 | 79.9 | 83.5 | 74.7 | 82.6 | 75.9 | 0.79 |
MRIv | 71.8 | 94.4 | 39.3 | 69.1 | 83.0 | 0.80 |
MRIv, Demographics, APOE4 | 73.3 | 95.8 | 40.9 | 70.0 | 87.2 | 0.83 |
Ensemble linear discriminant | ||||||
HipV | 83.2 | 87.1 | 77.5 | 84.7 | 80.7 | 0.90 |
HipV, Demographics, APOE4 | 85.6 | 90.9 | 77.7 | 85.2 | 85.1 | 0.90 |
MRIv | 90.8 | 94.4 | 85.5 | 90.3 | 91.4 | 0.95 |
MRIv, Demographics, APOE4 | 92.8 | 95.8 | 88.3 | 92.2 | 93.6 | 0.96 |
Boosted Trees | ||||||
HipV | 79.4 | 86.8 | 68.9 | 79.9 | 78.4 | 0.88 |
HipV, Demographics, APOE4 | 83.9 | 86.9 | 79.5 | 85.9 | 80.8 | 0.92 |
MRIv | 88.3 | 91.1 | 84.3 | 89.3 | 86.8 | 0.93 |
MRIv, Demographics, APOE4 | 88.5 | 91.6 | 83.9 | 89.1 | 87.5 | 0.93 |
Ensemble Random Forest | ||||||
HipV | 75.5 | 81.3 | 67.1 | 78.0 | 71.4 | 0.88 |
HipV, Demographics, APOE4 | 82.9 | 86.3 | 77.9 | 84.9 | 79.8 | 0.92 |
MRIv | 88.0 | 90.5 | 84.3 | 59.2 | 86.1 | 0.93 |
MRIv, Demographics, APOE4 | 88.5 | 90.8 | 85.1 | 89.8 | 86.5 | 0.94 |
HV, hippocampal volume; SVM, support vector machine; KNN, k-nearest neighbors; PPV, positive predictive value; NPV, negative predictive value; AUC, area under curve; HipV, hippocampal volume; MRIv, all MRI volumetrics.