Deep learning based on susceptibility-weighted MR sequence for detecting cerebral microbleeds and classifying cerebral small vessel disease

Table 2 Metrics in CSVD classification for the model and the neurologist group

Metrics	Internal test set	External test set
Metrics	Mean of 10-told cross-validation	Mean of 10-told cross-validation	Ensemble model	Neurologist group
Weighted-average AUC	0.865 (0.841–0.884)	0.899 (0.884–0.914)	0.908 (0.895–0.921)	–
Accuracy	0.732 (0.718–0.746)	0.717 (0.704–0.731)	0.819 (0.768–0.870)	0.759 (0.740–0.779)
Weighted-average precision	0.755 (0.743–0.768)	0.717 (0.702–0.733)	0.864 (0.831–0.897)	0.767 (0.747–0.786)
Weighted-average F1-score	0.733 (0.719–0.747)	0.705 (0.691–0.719)	0.829 (0.782–0.876)	0.762 (0.742–0.781)

The numbers in parentheses are the 95% confidence interval
CSVD cerebral small vessel disease, AUC area under the receiver operating characteristic curves

ISSN: 1475-925X