Sensitivity Analysis of Alzheimer’s Disease-Related Patterns

ORAL

Abstract

One of biomarkers for the Alzheimer’s disease (AD) is Alzheimer's Disease-Related Pattern (ADRP), identified with scaled subprofile model principal component analysis from fluorodeoxyglucose positron emission tomography (FDG-PET) brain images. For its clinical translation, it is important to understand how sensitive the ADRP is to variations in the analysis.

Sensitivity analysis was performed on 240 FDG-PET images (120 AD patients, 120 cognitive normal subjects (CN)). In each iteration, the ADRP was identified on random subpopulation with image preprocessing and pattern identification parameters randomly altered. Each identified ADRP was applied to a randomly selected validation group of 20 AD and 20 CN. The impact of different parameters on ADRP was assessed by calculating the effect size.

The effect size was consistently above 1, which suggests that the ADRP is able to differentiate between AD patients and CN subjects. The average effect size across tested parameters was 2.5, while reduced resolution resulted in the effect size around 1.5.

ADRP is a robust biomarker for Alzheimer’s disease, demonstrating insensitivity to variations in image preprocessing and parameter selection. These results support its reliability and potential for clinical application.

*This research was funded by Slovenian Research and Innovation Agency, grant number P1-0389.

Presenters

  • Eva Stokelj

    • University of Ljubljana

Authors

  • Urban Simoncic

    • University of Ljubljana
  • Eva Stokelj

    • University of Ljubljana