Skip to main content
. 2022 Jan 13;269(7):3511–3521. doi: 10.1007/s00415-021-10934-7

Table 3.

Decision tree based on various methods, clinical features and demographic characteristics

Decision tree Conditional attribute Predicted antibody (probability)
Psychiatric SPS spectrum Ataxia Dyskinesia Opsoclonus Chorea/choreoathetosis Dystonia Age Disease duration
Unsupervised method based on clinical feature only 0 0 1 0 GAD (0.47)
0 1 GAD (0.67)
0 0 1 1 ANNA-2/Ri (0.68)
1 NMDAR (0.71)
0 0 0 1 NMDAR (0.87)
0 0 0 0 1 CRMP5/CV2 (0.41)
Supervised method based on clinical features only 0 1 0 0 GAD (0.42)
1 0 GAD (0.61)
0 1 0 1 ANNA-2/Ri (0.40)
1 NMDAR (0.92)
Unsupervised method based on clinical features and demographic characteristics 1  ≥ 37 GAD (0.65)
0 1  < 37 GAD (0.76)
1 0 0  ≥ 37  < 0.23 NMDAR (0.38)
1 1  < 37 NMDAR (0.58)
1 0 0  ≥ 37  ≥ 0.23 LGI-1 (0.47)
0 1  ≥ 37 IgLON5 (0.42)
Supervised method based on clinical features and demographic characteristics 0 1 0  ≥ 37 GAD (0.44)
1  ≥ 37 GAD (0.65)
0 1  < 37 GAD (0.76)
1 1  < 37 NMDAR (0.52)
0 0 1  ≥ 37  < 0.05 CRMP5/CV2 (0.53)
0 1 1  ≥ 37 IgLON5 (0.67)