Functional Connectivity Correlation Clustering With MRI Harmonization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in using functional brain imaging for neurological/mental disorders is the small sample size leading to over-fitting and the inability to generalize classifiers across different imaging sites due to site-to-site differences in brain activity measurements.
Innovation Solution
A clustering device and method that uses supervised machine learning with under-sampling and sub-sampling, ensemble learning, and multiple co-clustering to generate a classifier model that corrects for measurement biases across multiple facilities, enabling accurate classification of brain functional connectivity correlation values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised machine learning is used to generate classifiers for brain functional connectivity data, then classification accuracy improves, but the model over-fits due to small sample size and cannot generalize across different imaging sites
Solution Approach 1:
The patent segments the training process into multiple stages: first training on local site data, then progressively incorporating data from other sites. This staged segmentation allows the model to learn site-specific patterns before generalizing, preventing over-fitting while maintaining classification accuracy.
Solution Approach 2:
The patent dynamically adjusts model parameters including regularization strength, learning rate, and data sampling ratios based on the number of available sites and samples. These parameter changes enable the model to adapt to varying data conditions, improving generalization while maintaining accuracy.
2Quantity of substance
If data from multiple imaging sites is collected to increase sample size, then statistical power improves, but measurement biases from different facilities reduce classification reliability
Solution Approach 1:
The patent introduces harmonization techniques as intermediary processing steps that transform data from different imaging sites into a unified representation. This intermediary transformation removes site-specific measurement biases while preserving the underlying biological signals, enabling reliable multi-site classification.
Solution Approach 2:
Instead of directly combining raw data from multiple sites, the patent inverts the approach by first identifying and removing site-specific biases through comparative analysis, then combining the corrected data. This inversion strategy prevents measurement inconsistencies from contaminating the final model.
3Loss of time
If conventional machine learning methods are used with small sample sizes, then training time is reduced, but the models fail to capture sufficient variability in brain activity patterns
Solution Approach 1:
The patent performs preliminary data harmonization and feature selection before main training, and uses efficient sampling strategies during training. These preliminary actions prepare the data in advance, reducing the computational burden during training while ensuring the model captures sufficient variability for robust classification.
Data Source
AI summary
A brain functional connectivity correlation value clustering device for clustering subjects having a prescribed attribute on the basis of brain measurement data obtained from a plurality of facilities, wherein a plurality of MRI devices capture resting state fMRI image data of a healthy cohort and a patient cohort; a computing system 300 performs generation of an identifier as ensemble learning of “supervised learning” between harmonized component values of correlation matrixes and disease labels of each of the subjects, selects, during the ensemble learning, features for clustering in accordance with importance from the features specified for generating an identifier for a disease label, and performs multiple co-clustering by “unsupervised learning.”