Skip to main content
. 2019 Aug 8;50(11):1852–1861. doi: 10.1017/S0033291719001934

Table 2.

Classification of patients with schizophrenia and healthy controlsa

Whole-brain images Functional connectivity Graph-based metrics
LR (%) SVM (%) DL (%) LR (%) SVM (%) DL (%) LR (%) SVM (%) DL (%)
Dataset 1
Accuracy 53.54 59.47 50.27 81.81 83.33 84.05 72.28 74.99 69.07
Sensitivity 38.13 52.75 32.64 69.23 100.00 73.63 58.46 62.75 66.04
Specificity 68.95 66.19 67.90 94.38 66.67 94.48 86.10 87.24 72.10
p valueb 0.155 0.056 0.421 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Dataset 2
Accuracy 50.00 50.00 50.00 80.01 77.14 77.97 65.28 71.21 68.84
Sensitivity 40.00 40.00 40.00 75.15 87.58 62.73 53.33 53.79 55.15
Specificity 60.00 60.00 60.00 84.87 66.70 93.22 77.24 88.63 82.54
p valueb 0.375 0.433 0.648 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Dataset 3
Accuracy 53.05 54.65 51.44 82.41 87.31 80.50 79.66 78.69 72.74
Sensitivity 58.67 58.67 42.89 71.11 100.00 75.11 73.56 65.33 65.33
Specificity 47.44 50.64 60.00 93.72 74.62 85.90 85.77 92.05 79.62
p valueb 0.180 0.088 0.777 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Dataset 4
Accuracy 51.10 54.37 53.74 81.19 79.62 81.69 66.82 63.81 65.00
Sensitivity 60.00 62.86 62.86 80.48 96.67 68.10 58.57 46.67 41.90
Specificity 42.21 45.88 44.63 81.91 62.57 95.29 75.07 80.96 88.09
p valueb 0.408 0.296 0.3182 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Dataset 5
Accuracy 56.00 56.00 54.52 79.44 81.32 80.96 57.19 71.33 67.41
Sensitivity 14.52 60.00 52.22 67.78 97.78 67.78 46.67 46.67 61.11
Specificity 86.67 52.00 56.81 91.10 64.86 94.14 67.71 96.00 73.71
p valueb 0.236 0.087 0.196 <0.001 <0.001 <0.001 0.019 <0.001 <0.001
Average Sensitivity 42.26 54.86 46.12 72.75 96.41 69.47 58.12 55.04 57.91
Average Specificity 61.05 54.94 57.87 89.20 67.08 92.61 78.38 88.98 79.21
Average Accuracy 52.74 54.90 51.99 80.97 81.74 81.03 68.25 72.00 68.61
a

Sensitivity and specificity were computed considering the patient group as the positive class.

b

Statistical significance was estimated using the permutation method (1000 permutations).

SVM, support vector machine; LR, logistic regression; DL, deep learning