Background. There is an increasing trend to represent domain knowledge in structured graphs, which provide efficient knowledgerepresentations for many downstream tasks. Knowledge graphs are widely used to model prior ...Background. There is an increasing trend to represent domain knowledge in structured graphs, which provide efficient knowledgerepresentations for many downstream tasks. Knowledge graphs are widely used to model prior knowledge in the form of nodesand edges to represent semantically connected knowledge entities, which several works have adopted into different medicalimaging applications. Methods. We systematically searched over five databases to find relevant articles that applied knowledgegraphs to medical imaging analysis. After screening, evaluating, and reviewing the selected articles, we performed a systematicanalysis. Results. We looked at four applications in medical imaging analysis, including disease classification, diseaselocalization and segmentation, report generation, and image retrieval. We also identified limitations of current work, such asthe limited amount of available annotated data and weak generalizability to other tasks. We further identified the potentialfuture directions according to the identified limitations, including employing semisupervised frameworks to alleviate the needfor annotated data and exploring task-agnostic models to provide better generalizability. Conclusions. We hope that our articlewill provide the readers with aggregated documentation of the state-of-the-art knowledge graph applications for medicalimaging to encourage future research.展开更多
基金the National Library of Medicine under Award No.4R00LM013001 and was also supported by Amazon Machine Learning Research Award 2020.
文摘Background. There is an increasing trend to represent domain knowledge in structured graphs, which provide efficient knowledgerepresentations for many downstream tasks. Knowledge graphs are widely used to model prior knowledge in the form of nodesand edges to represent semantically connected knowledge entities, which several works have adopted into different medicalimaging applications. Methods. We systematically searched over five databases to find relevant articles that applied knowledgegraphs to medical imaging analysis. After screening, evaluating, and reviewing the selected articles, we performed a systematicanalysis. Results. We looked at four applications in medical imaging analysis, including disease classification, diseaselocalization and segmentation, report generation, and image retrieval. We also identified limitations of current work, such asthe limited amount of available annotated data and weak generalizability to other tasks. We further identified the potentialfuture directions according to the identified limitations, including employing semisupervised frameworks to alleviate the needfor annotated data and exploring task-agnostic models to provide better generalizability. Conclusions. We hope that our articlewill provide the readers with aggregated documentation of the state-of-the-art knowledge graph applications for medicalimaging to encourage future research.
