Fig. 3: Model performance in predicting gait impairment severity among 93 participants within the training dataset. | npj Digital Medicine

Fig. 3: Model performance in predicting gait impairment severity among 93 participants within the training dataset.

From: Deep learning-enabled accurate assessment of gait impairments in Parkinson’s disease using smartphone videos

Fig. 3: Model performance in predicting gait impairment severity among 93 participants within the training dataset.

a The confusion matrix of the UPDRS score prediction. b The receiver operating characteristic (ROC) curve for each severity category, as well as the area under the ROC curve (AUC) and the micro-average AUC. c Performance metrics including macro precision, recall, specificity, F1 score and AUC. Except for recall, which remained constant, all other performance metrics exhibited slight improvements compared to those with the test dataset (Table 2 and Fig. 2). Notably, the AUC increased by 0.05.

Back to article page