Paper
Deep learning and artificial intelligence in radiology: Current applications and future directions
Published Nov 1, 2018 · K. Yasaka, O. Abe
PLoS Medicine
171
Citations
5
Influential Citations
Abstract
Radiological imaging diagnosis plays important roles in clinical patient management. Deep learning with convolutional neural networks (CNNs) is recently gaining wide attention for its high performance in recognizing images. If CNNs realize their promise in the context of radiology, they are anticipated to help radiologists achieve diagnostic excellence and to enhance patient healthcare. Here, we discuss very recent developments in the field, including studies published in the current PLOS Medicine Special Issue on Machine Learning in Health and Biomedicine, with comment on expectations and planning for artificial intelligence (AI) in the radiology clinic. Chest radiographs are one of the most utilized radiological modalities in the world and have been collected into a number of large datasets currently available to machine learning researchers. In this Special Issue, three groups of researchers applied deep learning to radiological imaging diagnosis using this modality. In the first, Pranav Rajpurkar and colleagues found that deep learning models detected clinically important abnormalities (e.g., edema, fibrosis, mass, pneumonia, and pneumothorax) on chest radiography, at a performance level comparable to practicing radiologists [1]. In a similar study, Andrew Taylor and colleagues developed deep learning models that detected clinically significant pneumothoraces on chest radiography with excellent performance on data from the same site—with areas under the receiver operating characteristic curve (AUC) of 0.94–0.96 [2]. Meanwhile, Eric Oermann and colleagues investigated how well deep learning models that detected pneumonia on chest radiography generalized across different hospitals. They found that models trained on pooled data from sites with different pneumonia prevalence performed well on new pooled data from these same sites (AUC of 0.93–0.94) but significantly less well on external data (AUC 0.75–0.89); additional analyses supported the interpretation that deep learning models diagnosing pneumonia on chest radiography are able to exploit confounding information that is associated with pneumonia prevalence [3]. Also in this Special Issue, Nicholas Bien and colleagues applied deep learning techniques to detect knee abnormalities on magnetic resonance (MR) imaging and found that the trained model showed near-human-level performance [4]. Taking these four studies together, we can interpret that deep learning is currently able to diagnose a number of conditions using radiological data, but such diagnostic models may not be robust to a change in location. These Special Issue studies join a growing number of applications of deep learning to radiological images from various modalities that can aid with detection, diagnosis, staging, and subclassification of conditions. Cerebral aneurysms can be detected on MR angiography with
Deep learning models can accurately diagnose various conditions on radiological images, but may not be robust to changes in location.
Full text analysis coming soon...