Information Sciences in Imaging 2012 Seminar Series: April


From Voxels to Meaning: Extraction and analysis of semantic information in medical images to improve reporting


The advent of digital radiology has allowed for much more efficient creation and interpretation of medical images. A result of this explosion in clinical imaging is that radiologists may now only contact patients/physicians through their reports. The importance of unambiguously communicating findings via reporting necessitates quality assurance of these radiological reports.

In this talk, I will present my preliminary work to measure correctness of findings described and completeness of information in the report. Using techniques from machine learning and sensitivity analysis, we can perform real-time analysis of semantic information in radiology reports. Furthermore, we present methods to request missing information from radiologists. Finally, we present an application that demonstrates the utility of semantically rich reports for content-based image retrieval.