IBM Watson Uses AI to Determine Patients Who Need Contrast

According to a September 18 paper in the Journal of Digital Imaging, the UCSF research team used IBM Watson’s language processing capacity to create an algorithm, which would automatically assign intravenous contrast for musculoskeletal MRI protocols. The aim was to develop and validate a machine learning-based natural language classifier capable of automatically assigning the use of intravenous contrast for musculoskeletal MRI protocols, based upon the study’s free-text clinical indication, thus improving efficiency of the protocoling radiologist and potentially reducing errors.

The researchers then compared the accuracy of IBM Watson, traditional computers, and a radiologist for determining which patients should receive contrast for musculoskeletal MRI scans based upon clinical indications. They found that Watson achieved an overall accuracy of 83.2% as compared to the original radiologist. In addition to its high accuracy in assigning MRI contrast, IBM Watson also offers the advantage of functioning without preprocessing, with the operator requiring only a minimal understanding of machine-learning fundamentals, according to the researchers.

The researchers concluded that the implementation of this automated MRI contrast determination system as a clinical decision support tool can save significant time and effort for the radiologist while potentially reducing error rates, and does not require any change in order entry or workflow.