Brain Functional Connectivity Harmonization via Bias Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in using brain functional connectivity data for neurological/mental disorder diagnosis is the difficulty in generalizing brain activity classifiers across different facilities due to site-to-site differences in measurement biases and small sample sizes, leading to over-fitting and poor performance.
Innovation Solution
A method and system using a generalized linear mixed model to adjust brain functional connectivity correlation values by calculating measurement biases at multiple sites, correcting these values to harmonize brain activity classifiers across facilities, and employing machine learning with feature selection to generate accurate and generalizable classifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If brain functional connectivity data is collected from multiple facilities, then the quantity of measurement data increases, but measurement biases and site-to-site differences worsen
Solution Approach 1:
The patent introduces a measurement bias correction model as an intermediary to mediate between the raw measurement data from multiple facilities and the final brain activity classifier. This model estimates and removes site-specific measurement biases, allowing data from multiple facilities to be combined while correcting for their respective biases, thus resolving the contradiction between data quantity and measurement precision.
Solution Approach 2:
The patent changes the parameters of the measurement data by adjusting correlation values based on estimated measurement biases. Specifically, it modifies the functional connectivity correlation values using correction factors derived from the bias model, transforming the raw measurements into bias-corrected values that can be reliably combined across facilities.
2Device complexity
If brain activity classifiers are trained with limited sample sizes, then the model complexity is reduced, but over-fitting increases and generalizability worsens
Solution Approach 1:
The patent implements a feedback mechanism where the measurement bias correction model continuously refines its estimates based on the training data and validation performance. The bias correction factors are adjusted based on the actual performance metrics, allowing the model to learn from its predictions and improve generalizability without requiring excessive training data or model complexity.
Solution Approach 2:
The patent performs preliminary bias correction on the measurement data before training the brain activity classifier. By pre-adjusting the data to remove site-specific biases, the subsequent classification task can be performed with reduced model complexity while achieving better generalizability, as the data is already prepared to minimize confounding effects.
3Adaptability or versatility
If measurement biases are corrected across multiple facilities, then the harmonization of brain activity measurements improves, but the computational complexity increases
Solution Approach 1:
The patent segments the bias correction process into separate, modular components: a measurement bias estimation module, a bias correction module, and a classification module. Each component operates independently on specific data subsets, allowing the computational complexity to be distributed and managed systematically rather than attempting to solve the entire problem at once.
Solution Approach 2:
The patent changes the parameters of the correlation values by applying correction factors that adjust for measurement biases. This transformation modifies the functional connectivity metrics in a way that harmonizes data across facilities while keeping the computational process manageable through efficient parameter updates rather than complete data reconstruction.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the harmonization of brain activity measurements across multiple facilities, improving the accuracy and generalizability of brain activity classifiers for neurological/mental disorder diagnosis, enabling objective determination of healthy or diseased states.
Implementation Method 1
The blood oxygen level dependent signal that reflects a brain activity measured by fMRI devices is referred to as a BOLD signal (Blood Oxygen Level Dependent Signal). The changes in blood stream cause changes in NMR signal intensity, since oxygenated hemoglobin and deoxygenated hemoglobin in the blood have different magnetic properties.
Implementation Method 2
The changes in blood stream cause changes in NMR signal intensity, since oxygenated hemoglobin and deoxygenated hemoglobin in the blood have different magnetic properties.
Implementation Method 3
using a generalized linear mixed model, calculating measurement bias for each element by of the brain functional connectivity matrix, as a fixed effect at each measurement site with respect to an average of the corresponding element across the plurality of measurement sites and across the plurality of traveling subjects
Implementation Method 4
employing machine learning with feature selection to generate accurate and generalizable classifiers
Data Source
Figure 1
Figure 2(A)~2(B)
Figure 3(A)~3(B)
AI summary
Brain Functional Connectivity Correlation Value Adjustment Method, Brain Functional Connectivity Correlation Value Adjustment System, Brain Activity Classifier Harmonization Method, Brain Activity Classifier Harmonization System, and Brain Activity Biomarker System A harmonization system for a brain activity classifier harmonizing brain measurement data obtained at a plurality of sites to realize a discrimination process based on brain functional imaging: obtains data, for a plurality of traveling subjects as common objects of measurements at each of the plurality of measurement sites, resulting from measurements of brain activities of a predetermined plurality of brain regions of each of the traveling subjects; calculates, for each of the traveling subjects, prescribed elements of a brain functional connectivity matrix representing the temporal correlation of brain activities of a set of the plurality of brain regions; using a generalized linear mixed model, calculates measurement bias data 3108 for each element of the brain functional connectivity matrix, as a fixed effect at each measurement site with respect to an average of the corresponding element across the plurality of measurement sites and across the plurality of traveling subjects; and thereby executes a harmonizing process.