Fig. 6.

LA-REIMS classification according to cervical disease severity. (a) Receiver operating characteristic (ROC) curve after employing random forest as the classification method. The generated values for the area under the curve (AUC) along with 95% confidence intervals (CI) are given within the plot. (b) Predicted class probabilities for each sample allowing visualisation of the misclassified samples (CIN2+ shown as black dots; normal controls shown as white dots). As a balanced subsampling approach is used for model training, the classification boundary is always at the centre (x = 0.5, the dotted line). (c) Confusion matrix showing the number of true positives (41/45), true negatives (40/55), false positives (15/55) and false negatives (4/45). Sensitivity and specificity are given in the green-highlighted regions, being 91% and 73% respectively.

CIN: cervical intraepithelial neoplasia; HPV: human papillomavirus.