Fig. 3From: Prediction of dysphagia aspiration through machine learning-based analysis of patients’ postprandial voicesThe pre-trained models demonstrated higher performance compared to the non-pre-trained models. Among the four models, the mn30_as (pre-trained model) performed the best on average in combined model and male model. However, for the female model, the mn4.0 (non-pre-trained model) was the best on average. The ROC curve was plotted, and the AUC (Area Under the Curve) was calculatedROC curve for each prediction model. Back to article page