RESUMO
Purpose: Building accurate and robust artificial intelligence systems for medical image assessment requires the creation of large sets of annotated training examples. However, constructing such datasets is very costly due to the complex nature of annotation tasks, which often require expert knowledge (e.g., a radiologist). To counter this limitation, we propose a method to learn from medical images at scale in a self-supervised way. Approach: Our approach, based on contrastive learning and online feature clustering, leverages training datasets of over 100,000,000 medical images of various modalities, including radiography, computed tomography (CT), magnetic resonance (MR) imaging, and ultrasonography (US). We propose to use the learned features to guide model training in supervised and hybrid self-supervised/supervised regime on various downstream tasks. Results: We highlight a number of advantages of this strategy on challenging image assessment problems in radiography, CT, and MR: (1) significant increase in accuracy compared to the state-of-the-art (e.g., area under the curve boost of 3% to 7% for detection of abnormalities from chest radiography scans and hemorrhage detection on brain CT); (2) acceleration of model convergence during training by up to 85% compared with using no pretraining (e.g., 83% when training a model for detection of brain metastases in MR scans); and (3) increase in robustness to various image augmentations, such as intensity variations, rotations or scaling reflective of data variation seen in the field. Conclusions: The proposed approach enables large gains in accuracy and robustness on challenging image assessment problems. The improvement is significant compared with other state-of-the-art approaches trained on medical or vision images (e.g., ImageNet).
RESUMO
BACKGROUND: Large databases of genetic data are often biased in their representation. Thus, selection of genetic data with desired properties, such as evolutionary representation or shared genotypes, is problematic. Selection on the basis of epidemiological variables may not achieve the desired properties. Available automated approaches to the selection of influenza genetic data make a tradeoff between speed and simplicity on the one hand and control over quality and contents of the dataset on the other hand. A poorly chosen dataset may be detrimental to subsequent analyses. RESULTS: We developed a tool, Tree Pruner, for obtaining a dataset with desired evolutionary properties from a large, biased genetic database. Tree Pruner provides the user with an interactive phylogenetic tree as a means of editing the initial dataset from which the tree was inferred. The tree visualization changes dynamically, using colors and shading, reflecting Tree Pruner actions. At the end of a Tree Pruner session, the editing actions are implemented in the dataset. Currently, Tree Pruner is implemented on the Influenza Research Database (IRD). The data management capabilities of the IRD allow the user to store a pruned dataset for additional pruning or for subsequent analysis. Tree Pruner can be easily adapted for use with other organisms. CONCLUSIONS: Tree Pruner is an efficient, manual tool for selecting a high-quality dataset with desired evolutionary properties from a biased database of genetic sequences. It offers an important alternative to automated approaches to the same goal, by providing the user with a dynamic, visual guide to the ongoing selection process and ultimate control over the contents (and therefore quality) of the dataset.