Fig. 1: Model performance evaluated using AUC and precision–recall curves. | npj Digital Medicine

Fig. 1: Model performance evaluated using AUC and precision–recall curves.

From: SpeechCARE: dynamic multimodal modeling for cognitive screening in diverse linguistic and speech task contexts

Fig. 1

a AUC-ROC curves (micro vs. weighted averaging). The micro average aggregates outcomes across all classes, yielding an AUC of 86.83%, while the weighted average, reflecting class frequency, achieves an AUC of 80.67%. b Precision–recall curves. The micro average achieves an average precision of 74.73%, and the weighted average yields an average precision of 73.50%. Overall, these metrics highlight how SpeechCARE balances sensitivity and specificity (in AUC) as well as precision and recall across the three diagnostic categories.

Back to article page