Table 4 Comparison of two step classification results using EC and EC-EO data for classification with 10-fold cross-validation. Classification was done using the k-nearest neighbour. Each classifier first identifies and filters HCs from non-healthy participants (part 1), and those identified as non-healthy are then carried through to the next stage of classification. Non-healthy individuals (part 2) are then classified within the multi-class environment based on which of the three dementia subgroups they are likely to belong to (AD, DLB, PDD)
From: Investigating the power of eyes open resting state EEG for assisting in dementia diagnosis
Classifier 1 (EC+EO) | ||||||
Accuracy | Specificity | CI Spec | Sensitivity | CI Sens | Weighted Average AUC | |
Step-1 HC against Non healthy | ||||||
HC | 87.70 ± 1.34 | 86.0 ± 7.3 | 0.10 | 93.3 ± 1.7 | 0.13 | 0.91 ± 0.04 |
Dementia Type | Step 2 - Separation of individual dementia groups | |||||
AD | 83.20 ± 1.10 | 89.47 ± 3.71 | 0.10 | 75.00 ± 0.76 | 0.28 | 0.83 ± 0.01 |
DLB | 85.60 ± 0.89 | 88.89 ± 1.48 | 0.13 | 90.48 ± 6.23 | 0.13 | 0.86 ± 0.01 |
PDD | 96.8 ± 1.10 | 90.09 ± 9.3 | 0.00 | 88.24 ± 2.64 | 0.16 | 0.98 ± 0.03 |
Classifier 2 (EC Only) | ||||||
Accuracy | Specificity | CI Spec | Sensitivity | CI Sens | Weighted Average AUC | |
Step-1 HC against Non healthy | ||||||
HC | 84.30 ± 2.29 | 73.30 ± 2.00 | 0.13 | 86.00 ± 5.57 | 0.18 | 0.88 ± 0.04 |
Dementia Type | Step 2 - Separation of individual dementia groups | |||||
AD | 81.60 ± 4.98 | 84.22 ± 7.46 | 0.13 | 58.3 ± 2.94 | 0.37 | 0.73 ± 0.04 |
DLB | 79.2 ± 1.79 | 81.46 ± 5.21 | 0.16 | 81.00 ± 3.53 | 0.19 | 0.86 ± 0.03 |
PDD | 96.4 ± 4.98 | 90.09 ± 9.3 | 0.00 | 88.2 ± 0.89 | 0.16 | 0.98 ± 0.03 |