Doing the math on quantitative imaging

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - Lisa Fratt - Portrait
Lisa Fratt, Editor

Quantitative imaging promises much, including greater precision and replicability and reduced variability. However, as with most phenomenal advances, caution is merited as new challenges lurk below the surface.  

Consider semiquantitative cardiac MR perfusion analysis software. The technology promises to streamline perfusion analysis and add a level of accuracy compared with visual analysis by a radiologist or cardiologist. But sometimes quantification, like math, can come back to haunt the user. In fact, that’s what Handayani et al found when the research team compared four software packages for cardiac MR perfusion analysis.

Although individual software packages provide reasonable intra- and interobserver reproducibility and repeatability, the researchers reported significant variability among results from different software packages.

Given the findings, Handayani and colleagues issued a list of recommendations for providers using these packages. These are:

  • Perform within group and/or sequential observations of quantitative perfusion parameters on one platform;
  • Specify and hold constant measurement settings;
  • Acknowledge that semiquantitative measurements require interpretation of values as estimates; and
  • Take into account the spectrum of the patient’s condition in clinical decision making.

Other quantitative imaging data are not plagued by these problems. Liver standardized uptake value normalized to lean body mass (SUV lean) on PET exams has emerged as a possible means of characterizing benign and malignant disease in staging exams and as a way to assess response to therapy.

Clinical application of this strategy hinges on understanding locational and interreader variability. Maya Viner, BA, from the department of radiology at Boston University School of Medicine, and colleagues evaluated studies interpreted by two different readers and samples in three different levels of the right lobe of the liver.

In this study, the researchers observed good interreader agreement at all three locations for SUV lean. These findings set the stage for further application of the measure in oncology practice.

On a larger level, the studies demonstrate the need to clearly define the capabilities, drawbacks and parameters for various applications of quantitative imaging.

How has your practice approached quantitative imaging? Please let us know.

Lisa Fratt, editor