Related Experiment Videos
Classifying Alzheimer's disease with brain imaging and genetic data using a neural network framework.
Kaida Ning1, Bo Chen2, Fengzhu Sun3
1USC Stevens Neuroimaging and Informatics Institute, Keck School of Medicine of University of Southern California, Los Angeles, CA, USA; Molecular and Computational Biology Program, University of Southern California, Los Angeles, CA, USA.
Neurobiology of Aging
|May 23, 2018
Summary
Neural networks effectively combine brain imaging and genetic data to accurately diagnose Alzheimer's disease (AD) and predict its progression. Key predictors include specific brain regions and the APOE gene, highlighting their role in AD risk.
Area of Science:
- Neuroimaging and Genetics
- Computational Neuroscience
- Biomedical Data Analysis
Background:
- Distinguishing Alzheimer's disease (AD) patients from cognitively normal (CN) individuals and predicting mild cognitive impairment (MCI) to AD progression remains a challenge.
- Integrating brain morphometric and genetic data offers a promising avenue for improved diagnostic and prognostic accuracy in AD.
Purpose of the Study:
- To develop and evaluate a neural network (NN) framework utilizing both structural brain imaging and genetic data for AD classification and prediction.
- To identify key imaging and genetic features, as well as their interactions, that contribute to AD risk and progression.
Main Methods:
- A neural network (NN) framework was applied to structural magnetic resonance imaging (MRI) and single nucleotide polymorphism (SNP) data from the AD Neuroimaging Initiative cohort.
- The dataset comprised 138 AD patients, 225 CN subjects, and 358 MCI patients.
- Model performance was assessed using the area under the receiver operating characteristic curve (AUC).
Main Results:
- NN models incorporating both brain imaging and SNP data significantly outperformed models using either data type alone for AD classification (AUC=0.992) and MCI to AD progression prediction (AUC=0.835).
- The most influential predictors identified were left middle temporal gyrus volume, left hippocampus volume, right entorhinal cortex volume, and the APOE ɛ4 risk allele.
- Novel interactions between specific brain regions and SNPs (rs10838725) associated with AD risk were uncovered.
Conclusions:
- Neural network models demonstrate high efficacy in classifying AD and predicting MCI to AD conversion when integrating multimodal data.
- The study successfully identified critical neuroimaging biomarkers and genetic factors, including APOE ɛ4, contributing to AD.
- This research underscores the potential of NN models for uncovering complex interactions underlying AD pathogenesis and informing risk assessment.