P4‐170: Combining Neurodegenerative Characterization With Amyloid Burden Measurement Using an Early Frame Amyloid PET Multivariate Classifier

Amyloid PET provides an enrichment tool for clinical trials and clinical diagnostic support. However, many amyloid+ early stage subjects do not worsen clinically during a clinical trial, and cognitive dysfunction can be driven by non-amyloid causes. A measure that characterizes neurodegeneration and is predictive of cognitive decline may provide a useful adjunct to amyloid measurement. Studies have shown correspondence between perfusion measured by early amyloid frames following tracer injection and FDG PET, using regions of interest. Multivariate machine learning approaches, by taking into account relationships between affected regions and maximizing signal/noise, may offer a more sensitive means for detection of disease related changes as we have demonstrated with FDG. Using summed dynamic florbetapir image frames acquired during the first six minutes post- injection for 104 ADNI subjects, we applied machine learning with iterative resampling to develop and test image classifiers measuring AD Progression. Training classes consisted of 10 NL amyloid-negative(-), 19 subjective memory complaints (SMC)-, 11 NL/SMC+, 9 MCI+, and 14 AD+ based upon clinical diagnosis and late timeframe amyloid status. Independent testing was applied through Leave-One-Out analysis and to 41 additional scans. Early frame amyloid (EFA) classification was compared to that of an independently developed FDG PET AD Progression classifier using FDG scans of the same subjects at the same time point. Correlations to clinical endpoints were compared. We also compared average Standardized Uptake Value Ratios in EFA scans to FDG and examined results when scoring EFA scans directly in the FDG classifier, using florbetapir scans as well as PiB scans. The EFA classifier produced a primary pattern similar to that of the FDG classifier (Figure 1) whose quantitative expression correlated with the FDG pattern (Figure 2; R-squared 0.71), and that within amyloid+ subjects (N=34) correlated with MMSE, CDR-sb, and ADAS-cog13 (R-squared 0.35, 0.37, 0.52). While highly correlated, there were regional differences between EFA scans and FDG scans that were addressed through the development of the EFA-specific classifier. These results show the ability to obtain a functional measure using EFA with the potential to achieve predictive utility approaching FDG through the use of multivariate classifier approaches. (a) FDG PET AD Progression classifier eigenimage and (b) Early frame amyloid AD Progression classifier image. Blue = hypometabolism (FDG) or hypoperfusion (EFA) and Red = preservation of metabolism (FDG) or perfusion (EFA), relative to whole brain. Classifier scores from (a) the early frame amyloid scans measured using the EFA functional classifier (Leave One Out independent test results) and (b) the FDG PET scans from the same subjects where available, using an independently developed FDG AD Progression classifier. (Column height = group mean, bars = SEM; number = number per group. NL = congnitively normal, SMC = normal with subjective memory complaint).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.