Fig. 3: Model performance on different datasets and tasks.
From: Automated abnormality classification of chest radiographs using deep convolutional neural networks
a Confusion matrices of VGG-19 CNN model performance on different datasets. Left: RSNA challenge dataset for normal versus abnormal classification. Middle: RSNA challenge dataset for normal versus lung opacity classification. Right: Indiana dataset for normal versus abnormal classification. b ROCs of CNN models’ performance on different datasets and tasks. Pre-train: CNNs pre-trained on the NIH “ChestX-ray 14” dataset for normal versus abnormal classification as weight initialization.
