Prevalence of radiographic findings between the ground truth cohort (manually annotated reports) and the Clever-Heart cohort.

For the ground truth corpus, the prevalence percentage is calculated based on known, manually annotated ground truths. For Clever-Heart, values are based on predictions from the CNN model.针对真实标注语料库（ground truth corpus），其患病率百分比通过已知的人工标注真实标签计算得出。针对Clever-Heart，其对应数值均基于卷积神经网络（Convolutional Neural Network, CNN）模型的预测结果生成。