Related Experiment Video
Updated: Jul 20, 2025
12:28
Abbiategrasso Brain Bank Protocol for Collecting, Processing and Characterizing Aging Brains
Published on: June 3, 2020
17.4K
Pathology steered stratification network for subtype identification in Alzheimer's disease
Enze Xu1, Jingwen Zhang1, Jiadi Li2
1Department of Computer Science, Wake Forest University, Winston-Salem, North Carolina, USA.
Medical Physics
|July 31, 2023
Summary
This study introduces a novel network to classify Alzheimer's disease (AD) subtypes using biological principles and neuroimaging. The approach identifies six distinct AD subtypes, aiding in early diagnosis and treatment strategies.
Area of Science:
- Neuroscience
- Computational Biology
- Medical Imaging Analysis
Background:
- Alzheimer's disease (AD) is a complex neurodegenerative disorder driven by beta-amyloid, tau pathology, and neurodegeneration.
- Effective late-stage treatments are lacking, emphasizing the need for early detection and prevention strategies.
- Current neuroimaging analysis methods for AD subtypes often neglect crucial pathological domain knowledge, potentially yielding inconsistent results.
Purpose of the Study:
- To develop a machine learning model integrating systems biology for clinical AD prognosis.
- To classify AD subpopulations based on biological principles, neurological patterns, and cognitive symptoms.
- To improve the accuracy and clinical relevance of AD subtype identification.
Main Methods:
- A novel pathology steered stratification network (PSSN) was developed, incorporating AD pathology knowledge via a reaction-diffusion model.
- The model considers non-linear biomarker interactions and brain network diffusion, predicting long-term individual progression trajectories from multimodal neuroimaging data.
- A deep predictive neural network was employed to analyze spatiotemporal dynamics, link clinical data, and assign individual subtype probabilities, with an evolutionary disease graph quantifying subtype transitions.
Main Results:
- The PSSN demonstrated superior performance in distinguishing between and within identified AD subtypes based on clinical scores.
- Application to aging populations revealed six distinct subtypes across the AD spectrum.
- Each identified subtype exhibited unique biomarker patterns correlating with specific clinical outcomes.
Conclusions:
- The PSSN effectively reduces neuroimage data, integrates pathological pathways (AT[N]-Net), predicts long-term biomarker evolution, and stratifies individuals into clinically relevant subtypes.
- This approach offers valuable insights for pre-symptomatic AD diagnosis and clinical treatment guidance.
- The methodology holds potential for generalization to other neurodegenerative diseases.
