Table 8 -.
Comparison between mean Dice Kappa (SI), detection/outline error rate (DER/OER), intra-class correlation (ICC), true positive rate (TPR), and positive prediction value (PPV) values of different classifiers for segmentation of WMHs using T1w, T2w, and PD data, ADC and ADNI1 datasets. Blue color indicates the best performance in terms of SI.
Dataset | ADC | ADNI1 | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Measure | SI | DER | OER | ICC | TPR | PPV | SI | DER | OER | ICC | TPR | PPV |
Naïve Bayes | 0.17±0.17* | 0.88±0.39 | 0.77±0.32 | 0.10 | 0.11 | 0.84 | 0.34±0.22 | 0.68±0.33 | 0.67±0.35 | 0.09 | 0.24 | 0.89 |
Logistic | 0.09±0.14* | 1.26±0.73 | 0.55±0.55 | 0.46 | 0.24 | 0.06 | 0.44±0.23 | 0.40±0.48 | 0.68±0.33 | 0.68 | 0.65 | 0.39 |
LDA | 0.28±0.21* | 0.08±0.05 | 1.35±0.42 | 0.45 | 0.24 | 0.66 | 0.48±0.28 | 0.17±0.34 | 0.82±0.53 | 0.62 | 0.46 | 0.73 |
QDA | 0.13±0.13* | 0.63±0.38 | 1.11±0.37 | 0.08 | 0.08 | 0.90 | 0.31±0.21 | 0.63±0.35 | 0.72±0.29 | 0.12 | 0.20 | 0.93 |
KNN | 0.28±0.24* | 0.77±0.68 | 0.66±0.34 | 0.69 | 0.44 | 0.22 | 0.59±0.23 | 0.25±0.33 | 0.56±0.27 | 0.74 | 0.67 | 0.58 |
Decision | 0.38±0.20* | 0.55±0.47 | 0.69±0.19 | 0.93 | 0.38 | 0.44 | 0.57±0.25 | 0.29±0.37 | 0.57±0.23 | 0.94 | 0.57 | 0.67 |
Random | 0.47±0.21* | 0.36±0.34 | 0.69±0.21 | 0.95 | 0.60 | 0.42 | 0.66±0.23 | 0.18±0.29 | 0.50±0.26 | 0.94 | 0.67 | 0.71 |
Bagging | 0.17±0.18* | 0.88±0.70 | 0.77±0.46 | 0.37 | 0.51 | 0.11 | 0.54±0.22 | 0.24±0.38 | 0.67±0.33 | 0.59 | 0.75 | 0.47 |
SVM | 0.31±0.21* | 0.62±0.54 | 0.76±0.31 | 0.65 | 0.48 | 0.31 | 0.61±0.24 | 0.25±0.43 | 0.56±0.34 | 0.83 | 0.62 | 0.70 |
AdaBoost | 0.44±0.21* | 0.43±0.47 | 0.69±0.21 | 0.93 | 0.53 | 0.42 | 0.64±0.24 | 0.20±0.33 | 0.55±0.32 | 0.94 | 0.66 | 0.73 |