Development brain aware graph neural network model for cognitive resilience in early Alzheimer’s Disease
- Abstract
- 알츠하이머병은 인지 기능의 점진적인 손상을 특징으로 하지만, 병리학적 변화에 대한 개인의 질병 양상에는 상당한 차이가 있습니다. 본 연구는 인지적으로 회복력이 있는 개인과 취약한 개인을 구별하는 구조적 연결성 바이오마커를 식별하여 병리학적 증상에도 불구하고 인지 기능을 유지하는 메커니즘에 대한 통찰을 제공합니다. 우리는 인지능력 회복력 있는 그룹과 취약한 그룹 간의 구조적 연결성 차이를 분석하기 위해 베이지안 가설 검정을 사용하여, 인지 회복력과 관련된 중요한 특징들을 밝혀냈습니다. 또한,뇌연결성에대한도메인지식을통합하여임상점수예측정확도를향상시키는그 래프 신경망을 위한 새로운 뇌 특화 리드아웃 레이어를 제안합니다. 실험 결과는 제안된 그래프 신경망 모델이 신경영상 데이터에서 효과적이고 더 높은 해석을 보여주었으며, 신경퇴행성질환의조기발견및표적개입의가능성을강조합니다.우리의연구결과는 인지 회복력에 대한 더 깊은 이해를 제공하며, 알츠하이머병의 진단 및 치료를 향상시키 기 위한 새로운 가능성을 제시합니다.|Alzheimer’s disease (AD) is characterized by progressive impairment of cognitive functions, but significant variability exists in individual susceptibility to neuropathological changes. This study aims to identify structural connectivity biomarkers that distinguish cognitively resilient individuals from those who are vulnerable, providing insights into mechanisms sustaining cognitive performance despite AD pathology. We use Bayesian hypothesis testing to analyze differences in structural connectivity between resilient and vulnerable groups, revealing significant features associated with cognitive resilience. Furthermore, we propose a novel brain-aware readout layer for Graph Neural Networks (GNNs), incorporating domain-specific knowledge of brain connectivity to improve prediction accuracy for clinical scores. Experimental results demonstrate the effectiveness and interpretability of the proposed GNN model in neuroimaging data, highlighting its potential for early detection and targeted intervention in neurodegenerative conditions. Our findings contribute to a deeper understanding of cognitive resilience and present new avenues for enhancing diagnosis and treatment of Alzheimer’s disease.
- Issued Date
- 2025
- Type
- Thesis
- Department
- 대학원 AI대학원
- Advisor
- Kim, Mansu
- Table Of Contents
- Abstract (English) i
Abstract (Korean) iv
List of Contents v
List of Tables vii
List of Figures viii
1 Introduction 1
1.1 Introduction of Alzheimer’s disease 1
1.2 Resilience in cognitive functions 2
1.3 Research objectives 3
2 Related works 5
2.1 Hypothesis Testing 5
2.1.1 Frequentist Hypothesis Testing 5
2.1.2 Bayesian Hypothesis Testing 6
2.2 Graph Neural Networks 7
2.2.1 Aggregation Layer 7
2.2.2 Readout Layer 10
3 Biomarkers for cognitive resilience 12
3.1 Background 12
3.2 Material and Methods 13
3.2.1 Participants 13
3.2.2 Neuropsychological testing 14
3.2.3 MRI image acquisition and processing 14
3.2.4 [18F]-flutemetamol PET image acquisition and processing 15
3.2.5 Between group difference based on Bayesian Statistics 16
3.2.6 Assessing the Predictive Power of Identified SC 17
3.3 Results 17
3.3.1 Baseline demographic and clinical data 17
3.3.2 Between-group Differences in SC 18
3.3.3 Prediction performance of identified connectomics biomarker 20
3.4 Conclusion 21
4 Brain-Aware Readout Layers in GNNs 22
4.1 Background 22
4.2 Material and Methods 23
4.2.1 Participants 23
4.2.2 Data Preprocessing 23
4.2.3 Brain-aware Graph Neural Network 24
4.3 Experiments and results 28
4.3.1 Experimental Setups 28
4.3.2 Performance Comparison Across Readout Layer on Various GNN
Model 29
4.3.3 Effectiveness of Prior Knowledge in BA Readout Layer 30
4.3.4 Interpretability of the GNN Model with BA Readout Layer 30
4.3.5 Conclusion 33
5 Discussion 36
Summary 37
Acknowledgements 48
- Degree
- Master
- Appears in Collections:
- Department of AI Convergence > 3. Theses(Master)
- 파일 목록
-
-
관련 파일이 존재하지 않습니다.
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.