Joeky Senders

151 Deep Learning and NLP learning curves ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ●● ● ●● A 0.5 0.6 0.7 0.8 0.9 1.0 0 500 1000 1500 2000 2500 3000 Sample size Mean bootstrapped AUC ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● B 0.90 0.92 0.94 0.96 0.98 1.00 0 200 400 600 800 1000 1200 1400 Sample size Mean bootstrapped AUC Model ● ● ● ClinicalTextMiner Lasso regression Logistic regression FIGURE 1. Incremental model performance comparing ClinicalTextMiner to regression-based algorithms according to the area under the receiver operating characteristic curve (A). Enlarged panel can be found on the right (B). The dashed line represents the 0.95 performance threshold and the dotted line represents the 0.98 performance threshold. Abbreviations: AUC= area under the receiver operating characteristic curve. ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● A 0.5 0.6 0.7 0.8 0.9 1.0 0 500 1000 1500 2000 2500 3000 Sample size Mean bootstrapped AUC ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● B 0.90 0.92 0.94 0.96 0.98 1.00 0 200 400 600 800 1000 1200 1400 Sample size Mean bootstrapped AUC Model ● ● ClinicalTextMiner CNN FIGURE 2. Incremental model performance comparing ClinicalTextMiner to the best performing convolutional neural network (CNN) model according to the area under the receiver operating characteristic curve (A). Enlarged panel can be found on the right (B). The dashed line represents the 0.95 performance threshold and the dotted line represents the 0.98 performance threshold. Abbreviations: AUC= area under the receiver operating characteristic curve; CNN=convolutional neural network.