Premium
IC‐P‐128: Use of PONS as a normalizing region for [c‐11]pib‐pet scans: Effect on subject classification
Author(s) -
Lopresti Brian,
Klunk William,
Bi Wenzhu,
Cohen Ann,
Mathis Chester,
Price Julie
Publication year - 2011
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2011.05.093
Subject(s) - nuclear medicine , pons , outlier , medicine , psychology , mathematics , statistics
Background: Negligible levels of fibrillar Abeta deposits in the cerebellum (CER) in sporadic Alzheimer’s disease (AD) supports the use of CER as reference for normalizing regional [C]PiB retention measures. A CER reference, however, may not be ideal for all groups, particularly eoFAD or Down subjects, where significant CER Abeta deposition is a common autopsy finding. For this reason, we examined the suitability of pons (PON) as a reference region.Methods: [C]PiB PETwas performed in 190 subjects (93 Control, 51 MCI, 46 AD) and 46 of these (21 Control, 14, 11 AD) underwent fully quantitative imaging involving 90 min dynamic scanning, arterial input function determination, and estimation of Logan DVR outcomes (ART90). Regional SUVR50-70 retentionmeasureswere determined for all using CER (SUVRCER) and PON (SUVRPON) as reference. For the quantitative subjects, results of the SUVRmethods were compared to ART90. Iterative outlier and k-means clustering approaches were used to classify subjects as PiB positive or negative. Results: CER:plasma ratios reached a plateau of w6 at 30 min, while PON:plasma plateau was slower and more transient, peaking at w12 at 50 min. Across cortical areas, SUVRCER was less biased and more highly correlated with ART90 than SUVRPON. However, both provided similar Cohen’s effect sizes for the distinction of AD and PIB negative Control groups. This is attributable to lower variance in SUVRPON outcomes despite a compressed dynamic range. Using the iterative outlier method, SUVRPON identified significantly more PIB + Cons than SUVRCER (38/93 vs. 27/93), while the overall rate of discordance between SUVRCER and SUVRPON was 12% (23/190). When available, visual interpretation agreed with theSUVRCER classification in 89% of the discordant cases. Use of a k-means clustering algorithm resulted in less discordance between SUVRPON and SUVRCER classification (7%), but classified fewer Control subjects as PiB + compared to the iterative outlier method using both SUVRCER and SUVRPON outcomes (17/ 93 and 21/93, respectively). Conclusions: SUVRPON effectively discriminates amyloid negative controls fromAD subjects andmay be useful when cerebellar data is unavailable or contraindicated. However, the transient kinetics underlying SUVRPON may lead to less reliable subject classification at the PiB + /-interface, relative to SUVRCER.