Interpretable Ensemble 3DCNN for Longitudinal sMRI Biomarker Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning methods for analyzing structural magnetic resonance imaging (sMRI) images in Alzheimer's disease focus on classification tasks and fail to effectively study longitudinal trajectory changes and spatial-temporal associations of neurodegenerative brain regions.
Innovation Solution
A method using an interpretable ensemble 3DCNN to extract a neuroimaging biomarker by preprocessing sMRI images, segmenting them into small cubes, and converting prediction probabilities into a multilevel P-score to analyze neurodegenerative changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning methods using independent training strategies and voting ensemble rules are used, then classification accuracy is improved, but the ability to study longitudinal trajectory changes and spatial-temporal associations is lost
Solution Approach 1:
The method segments the sMRI image into multiple small cubes (e.g., 25×25×25 voxels) that overlap in space and time. Each small cube is independently analyzed by base classifiers, but the segmentation strategy ensures temporal overlap between consecutive cubes, preserving longitudinal trajectory information while maintaining the benefits of independent classification.
Solution Approach 2:
The patent implements a hierarchical ensemble structure where multiple base classifiers (3DCNN models) are nested within a meta-classifier framework. The base classifiers process individual small cubes, and their predictions are integrated by the meta-classifier to produce final classification results, allowing both independent analysis and integrated interpretation of longitudinal patterns.
2Measurement precision
If PET imaging is used to acquire brain images, then diagnostic information quality is improved, but cost and accessibility deteriorate
Solution Approach 1:
The patent changes the imaging modality parameter from PET to sMRI, accepting lower intrinsic image quality in exchange for significantly reduced cost and improved accessibility. The method compensates for this parameter change by using advanced preprocessing techniques and ensemble 3DCNN analysis to extract meaningful neuroimaging biomarkers from the lower-cost sMRI data.
3Productivity
If sMRI image analysis focuses on classification tasks, then diagnostic efficiency is improved, but the ability to analyze progression patterns and neurodegenerative changes deteriorates
Solution Approach 1:
The patent makes the sMRI analysis system multi-functional by enabling it to perform both classification tasks (diagnostic efficiency) and progression pattern analysis (neurodegenerative changes). The ensemble 3DCNN framework with overlapping small cubes can simultaneously provide diagnostic classification and track longitudinal changes in neuroimaging biomarkers, serving multiple purposes from a single analysis pipeline.
Data Source
AI summary
The present invention provides a method for extracting a neuroimaging biomarker based on an interpretable ensemble three-dimensional convolutional neural network (3DCNN) to address limitations in the prior art. The present invention derives a novel neuroimaging biomarker P-score from prediction results obtained by an ensemble three-dimensional convolutional neural network model. The solution can help researchers to conduct studies on longitudinal trajectory changes of structural magnetic resonance imaging (sMRI) during the progression of Alzheimer's disease, and analyze an association of the longitudinal trajectory changes with neurodegenerative changes of Alzheimer's disease subjects. The extracted neuroimaging biomarker can provide a basis for predicting a sequence of intervention of brain regions in the neurodegenerative changes of Alzheimer's disease patients and upcoming clinical symptoms.