Related Experiment Videos
Deep Learning in Alzheimer's Disease: Diagnostic Classification and Prognostic Prediction Using Neuroimaging Data
Taeho Jo1,2,3, Kwangsik Nho1,2,3, Andrew J Saykin1,2,3
1Department of Radiology and Imaging Sciences, Center for Neuroimaging, Indiana University School of Medicine, Indianapolis, IN, United States.
Frontiers in Aging Neuroscience
|September 5, 2019
Summary
Deep learning shows promise for early Alzheimer's disease (AD) detection using neuroimaging. Combining deep learning with multimodal data significantly improves diagnostic accuracy for AD and mild cognitive impairment conversion.
Area of Science:
- Artificial Intelligence
- Neuroscience
- Medical Imaging
Background:
- Deep learning excels at analyzing complex, high-dimensional data, particularly in computer vision.
- Advancements in neuroimaging generate large datasets, driving interest in deep learning for Alzheimer's disease (AD) detection.
- Early and automated classification of AD is crucial for timely intervention.
Approach:
- A systematic review of deep learning applications in AD diagnosis using neuroimaging data was conducted.
- Publications from January 2013 to July 2018 were identified via PubMed and Google Scholar.
- Studies were evaluated and classified by algorithm type and neuroimaging modality.
Key Points:
- Deep learning models achieved high accuracy in AD classification (up to 98.8%) and mild cognitive impairment (MCI) to AD conversion prediction (up to 83.7%).
- Hybrid approaches combining traditional machine learning with stacked auto-encoders (SAE) for feature selection showed strong performance.
- Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) using raw neuroimaging data also yielded significant accuracies (up to 96.0% for AD classification).
- Optimal classification performance was achieved by integrating multimodal neuroimaging with fluid biomarkers.
Conclusions:
- Deep learning demonstrates significant potential for the diagnostic classification of AD using multimodal neuroimaging data.
- Ongoing research focuses on enhancing deep learning models by incorporating diverse data types (e.g., -omics) and improving transparency with explainable AI.