Building Disease Detection Algorithms with Very Small Numbers of Positive Samples
This addresses the challenge of limited annotated data for medical image analysis, particularly for rare diseases, though it is incremental as it builds on existing transfer learning and segmentation methods.
The authors tackled the problem of disease detection in medical images with very few positive samples by training a segmentation model on normal images and transferring its feature maps to a classification network, achieving detection of pericardial effusion and cardiac septal defects with as little as one positive sample per 17 negatives.
Although deep learning can provide promising results in medical image analysis, the lack of very large annotated datasets confines its full potential. Furthermore, limited positive samples also create unbalanced datasets which limit the true positive rates of trained models. As unbalanced datasets are mostly unavoidable, it is greatly beneficial if we can extract useful knowledge from negative samples to improve classification accuracy on limited positive samples. To this end, we propose a new strategy for building medical image analysis pipelines that target disease detection. We train a discriminative segmentation model only on normal images to provide a source of knowledge to be transferred to a disease detection classifier. We show that using the feature maps of a trained segmentation network, deviations from normal anatomy can be learned by a two-class classification network on an extremely unbalanced training dataset with as little as one positive for 17 negative samples. We demonstrate that even though the segmentation network is only trained on normal cardiac computed tomography images, the resulting feature maps can be used to detect pericardial effusion and cardiac septal defects with two-class convolutional classification networks.