Related Experiment Videos
Short-Duration [18F]florbetapir PET Reconstructions: Centiloid Agreement and Temporal-Window Analysis
Yusuke Kawasaki1, Kota Yokoyama2, Hirofumi Yamada1
1From the Department of Diagnostic Radiology (Y.K., K.Y., H.Y., M.K., J.O., J.T., U.T.), Radiology Center (K.T.), Neurology and Neurological Science (N.S., Y.N., M.O., M.Y., S.M.), Institute of Science Tokyo, Bunkyo-ku, Tokyo, Japan; Department of Neurology (N.S., M.Y.), Kudanzaka Hospital, Chiyoda-ku, Tokyo, Japan and Neurology (Y.N.), Saitama Prefectural Rehabilitation Center, Ageo, Saitama, Japan.
Background And Purpose:
Shorter [18F]florbetapir PET acquisitions could improve workflow, but the effects of reconstruction duration and temporal-window position on Centiloid values are not well characterized. We evaluated quantitative agreement across short-duration reconstructions, temporal-window effects, and visual agreement.
Materials And Methods:
This retrospective study included a primary cohort of 32 patients with early 1-, 2-, and 5-min reconstructions and a 10-min reference. A temporal-window cohort of 22 patients underwent early and late 5-min analysis. A validation cohort of 37 examinations underwent early, centered, and late 1- and 5-min reconstructions and a 10-min reference. Two readers independently classified early 1-min, early 5-min, and 10-min images across three separate reading sessions. Centiloid agreement was assessed using paired differences, Bland-Altman limits of agreement, and categorical agreement.
Results:
In the primary cohort, mean differences from the 10-min reference were -3.70, -2.97, and -1.65 CL for early 1-, 2-, and 5-min reconstructions, respectively. In the temporal-window cohort, late 5-min values averaged 2.86 CL higher than early values. The validation cohort reproduced both findings: early-to-late shifts were +5.53 CL at 1 min and +3.07 CL at 5 min. In the reader study, agreement with 10 min was 70.3%-86.5% at 1 min and 94.6%-100% at 5 min; interreader κ was 0.49, 0.77, and 0.89 at 1, 5, and 10 min, respectively.
Conclusions:
Early short-duration reconstructions yielded systematically lower Centiloid values, while later temporal windows yielded higher values at fixed duration. Visual agreement with 10 min was high at 5 min but lower at 1 min. Reconstruction duration and frame timing should be standardized before abbreviated protocols are considered substitutes for standard 10-min imaging.