Ameya Nayate MD, Presenter: Grant, Siemens AG

Jacob G. Dubroff MD, PhD, Abstract Co-Author: Research Grant, Siemens AG Speaker, Siemens AG Research Grant, General Electric Company Research Grant, Pfizer Inc Research Grant, Johnson & Johnson Research Grant, Functional Neuromodulation Ltd Research Grant, Bayer AG

James Eric Schmitt MD, PhD, Abstract Co-Author: Nothing to Disclose

Rekha Indira Kishore MD, Abstract Co-Author: Stockholder, General Electric Company

Ilya Michael Nasrallah MD, PhD, Abstract Co-Author: Nothing to Disclose

David A. Mankoff MD, PhD, Abstract Co-Author: Nothing to Disclose

Daniel Pryma MD, Abstract Co-Author: Speaker, Siemens AG Research Grant, Siemens AG Research Grant, Molecular Insight Pharmaceuticals, Inc Consultant, Molecular Insight Pharmaceuticals, Inc

[F-18] Florbetapir is a β-amyloid plaque imaging agent that specifically binds to cortical fibrillar -amyloid. However, there has been concern about consistency of the interpretation of Florbetapir images including concerns expressed by the FDA during its approval process. Semi-quantative analysis of Florbetapir images compares differential Florbetapir binding between the cerebrum and cerebellum, as a site not prone to amyloid deposition, expressed as cerebral–to–whole-cerebellar standard uptake value ratios (SUVr). Prior studies using SUVr have demonstrated high correlation between Florbetapir-PET and subsequent immunohistochemistry measurements of β amyloid. Therefore we hypothesized that incorporating SUVr into F-18 Florbetapir PET brain scan interpretation would reduce the inter-reader variability.

29 patients enrolled in the Alzheimer's disease neuroimaging initiative (ADNI 2) were included. Readers classified each case using a binary system, positive or negative for significant beta amyloid deposition. Each case was interpreted twice by each reader, once qualitatively and once with the aid of SUVr measurements generated by Scenium software (Siemens Medical). Cases were randomly assigned to 4 reading sessions separated by a washout period and interpreted by 5 blinded, board certified and Florbetapir-interpretation trained readers. No case was repeated within an individual session. To quantify inter-rater agreement, a kappa coefficient was calculated for the raters with and without the use of Scenium.

When Florbetapir PET brain studies were read qualitatively, there was inter-reader disagreement in 8/29 cases. When the same Florbetapir PET studies were read with SUVr, there was inter-reader disagreement for only 1 case. The kappa coefficient for the studies read with SUVr (0.94) was statistically significantly higher compared to the qualitatively only read studies (0.71), p < 0.005.

The use of semi-quantitative indices (SUVr) to aid the interpretation of Florbetapir images improves inter-reader agreement. Further study is needed to confirm the impact on the accuracy of interpretation.

Computer aided diagnostic software can decrease inter-reader variability of F-18 Florbetapir PET brain scan interpretation.

Nayate, A, Dubroff, J, Schmitt, J, Kishore, R, Nasrallah, I, Mankoff, D, Pryma, D, Does Computer Aided Diagnostic (CAD) Software Decrease Inter-reader Variability of Florbetapir PET Brain Scan Interpretation?.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13013752.html