FDG-PET Detects Amygdala-Hippocampal Connection Volume and Heterogeneity in Alzheimer's Disease Progression
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Objective This study delves into the potential of radiomics to identify the Alzheimer's disease (AD) stages and monitor its progression using FDG PET images. The particular objective is to explore the potential of tracking progression using a reduced and meaningful set of imaging biomarkers at the border of the Hippocampus and Amygdala. Method Our study utilized 18F-FDG PET scans from 513 participants spanning three stages of Alzheimer’s disease (AD) from the ADNI database. The hippocampus, amygdala, and entorhinal cortex consistently emerged as key regions of interest. To further investigate their interconnectivity, we defined the hippocampus-amygdala connecting region using a distance transform approach. Next, we systematically evaluated eight feature selection techniques in combination with six classification models to determine the most effective predictive framework. Finally, we conducted a Pearson correlation analysis to pinpoint the most significant features for AD classification. Results The connectivity area between the hippocampus and amygdala demonstrated superior efficacy for diagnosing AD. Two features shape_MeshVolume_Right, gldm_SmallDependenceLowGrayLevelEmphasis_left, glrlm_ShortRunLowGrayLevelEmphasis_left for this single region, could predict AD versus Control Normal (CN) individuals with ROC AUC = 0.91, and two features predict MCI versus AD with ROC AUC = 0.80, and a feature shape_LeastAxisLength_left, glszm_LargeAreaEmphasis_left for CN versus MCI with ROC AUC = 0.69. The features' mean values were able to demonstrate the incremental deterioration between groups of consecutive AD stages with statistical significance (p < 0.05). Conclusions In this study, we identify significant radiomic features within the hippocampus-amygdala connecting region and their utility in tracking AD progression. By limiting these specific biomarkers, we offer a simple but clinically relevant approach to AD diagnosis and monitoring and demonstrate the role of radiomics in early detection and intervention.