Subregional Biomarkers in FDG PET for Alzheimer’s Diagnosis and Staging: An Interpretable and Explainable model DOI Creative Commons
Ramin Rasi, Albert Güveniş

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 27, 2024

Abstract Objective To investigate the radiomics features of hippocampus and amygdala subregions in FDG-PET images that can best differentiate Mild Cognitive Impairment (MCI), Alzheimer’s Disease (AD), healthy patients. Methods Baseline data from 555 participants ADNI dataset were analyzed, comprising 189 cognitively normal (CN) individuals, 201 with MCI, 165 AD. The segmented based on DKT-Atlas, additional subdivisions guided by probabilistic atlases Freesurfer. Then radiomic (n=120) extracted 38 hippocampal 18 nuclei using PyRadiomics. Various feature selection techniques, including ANOVA, PCA, Chi-square, LASSO, applied alongside nine machine learning classifiers. Results Multi-Layer Perceptron (MLP) model combined LASSO demonstrated excellent classification performance: ROC AUC 0.957 for CN vs. AD, 0.867 MCI 0.782 MCI. Key regions, accessory basal nucleus, presubiculum head, CA4 identified as critical biomarkers. Features GLRLM (Long Run Emphasis) Small Dependence Emphasis (GLDM) showed strong diagnostic potential, reflecting subtle metabolic microstructural changes often preceding anatomical alterations. Conclusion Specific their four found to have a significant role early diagnosis its staging, severity assessment capturing shifts patterns. Furthermore, these offer potential insights into disease’s underlying mechanisms interpretability.

Language: Английский

Uncovering atrophy progression pattern and mechanisms in individuals at risk of Alzheimer's disease DOI Creative Commons
Christina Tremblay, Shady Rahayel, Alexandre Pastor‐Bernier

et al.

Brain Communications, Journal Year: 2025, Volume and Issue: 7(2)

Published: Jan. 1, 2025

Abstract Alzheimer's disease is associated with pre-symptomatic changes in brain morphometry and accumulation of abnormal tau amyloid-beta pathology. Studying the development prior to symptoms onset may lead early diagnostic biomarkers a better understanding pathophysiology. pathology thought arise from combination protein spreading via neural connections, but how these processes influence atrophy progression phases remains unclear. Individuals family history (FHAD) have an elevated risk disease, providing opportunity study phase. Here, we used structural MRI three databases (Alzheimer's Disease Neuroimaging Initiative, Pre-symptomatic Evaluation Experimental or Novel Treatments for Alzheimer Montreal Adult Lifespan Study) map FHAD assess constraining effects connectivity on progression. Cross-sectional longitudinal data up 4 years were perform analysis compared controls. PET radiotracers also quantify distribution isoforms at baseline. We first derived cortical maps using deformation-based 153 FHAD, 156 116 controls similar age, education sex next examined spatial relationship between patterns aggregates plaques deposition, neurotransmitter receptor transporter distributions. Our results show that there notably cingulate, temporal parietal cortices, more widespread severe disease. Both tended accumulate regions structurally connected The pattern its aligned existing FHAD. In our findings suggest propagation occurred earlier, previously intact connectome. Moreover, was found serotonin current demonstrates showing present specific cellular characteristics, uncovering some mechanisms involved pre-clinical clinical neurodegeneration.

Language: Английский

Citations

0
Subregional Biomarkers in FDG PET for Alzheimer’s Diagnosis and Staging: An Interpretable and Explainable model DOI Creative Commons
Ramin Rasi, Albert Güveniş

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 27, 2024

Abstract Objective To investigate the radiomics features of hippocampus and amygdala subregions in FDG-PET images that can best differentiate Mild Cognitive Impairment (MCI), Alzheimer’s Disease (AD), healthy patients. Methods Baseline data from 555 participants ADNI dataset were analyzed, comprising 189 cognitively normal (CN) individuals, 201 with MCI, 165 AD. The segmented based on DKT-Atlas, additional subdivisions guided by probabilistic atlases Freesurfer. Then radiomic (n=120) extracted 38 hippocampal 18 nuclei using PyRadiomics. Various feature selection techniques, including ANOVA, PCA, Chi-square, LASSO, applied alongside nine machine learning classifiers. Results Multi-Layer Perceptron (MLP) model combined LASSO demonstrated excellent classification performance: ROC AUC 0.957 for CN vs. AD, 0.867 MCI 0.782 MCI. Key regions, accessory basal nucleus, presubiculum head, CA4 identified as critical biomarkers. Features GLRLM (Long Run Emphasis) Small Dependence Emphasis (GLDM) showed strong diagnostic potential, reflecting subtle metabolic microstructural changes often preceding anatomical alterations. Conclusion Specific their four found to have a significant role early diagnosis its staging, severity assessment capturing shifts patterns. Furthermore, these offer potential insights into disease’s underlying mechanisms interpretability.

Language: Английский

Citations

0