Resting-State fMRI Brain Mapping With 3DCNN for Shorter MRI Scans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for brain functional mapping using resting state fMRI require lengthy scanning sessions, which are inconvenient for patients and inefficient for clinical applications, and existing analysis techniques are signal-to-noise limited with limited sensitivity and specificity.
Innovation Solution
A method utilizing a deep learning approach with a three-dimensional convolutional neural network (3DCNN) to analyze reduced amounts of resting state fMRI data, enabling accurate mapping of brain networks by calculating voxel probabilities and generating functional connectivity maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional RS-fMRI analysis methods are used, then measurement precision is limited due to signal-to-noise constraints, but scanning time is reduced
Solution Approach 1:
The patent replaces conventional signal processing methods (independent component analysis, seed-based correlation) with a deep learning-based artificial neural network system. The 3DCNN model processes fMRI data to identify resting state networks, substituting traditional mechanical signal analysis with intelligent pattern recognition that achieves superior measurement precision even with reduced scanning time.
Solution Approach 2:
The patent changes the fundamental parameter of data acquisition by training the 3DCNN model on datasets with varying time lengths. The model learns to extract meaningful functional mapping information from reduced-time fMRI scans, transforming the parameter relationship between scan duration and mapping accuracy in favor of shorter scanning times while maintaining high precision.
2Measurement precision
If longer scanning sessions are used, then functional mapping accuracy improves, but patient comfort and clinical efficiency deteriorate
Solution Approach 1:
The deep learning-based 3DCNN system replaces conventional analysis methods that require extensive data averaging over long periods. The neural network's ability to learn complex patterns from limited data enables accurate network characterization during shorter, more patient-friendly scanning sessions.
Solution Approach 2:
The patent performs preliminary training of the 3DCNN model using large datasets obtained from multiple subjects during research-phase scanning sessions. This preliminary action allows the model to learn robust features that can then be applied to individual patient scans with reduced time requirements, transferring knowledge from extensive training data to time-constrained clinical applications.
3Reliability
If conventional analysis methods are used, then sensitivity and specificity are limited, but system complexity is lower
Solution Approach 1:
The patent substitutes conventional signal processing algorithms with a deep learning-based 3DCNN system. This substitution significantly improves sensitivity and specificity in identifying resting state networks, despite the increased computational complexity. The neural network architecture with multiple convolutional layers and pooling operations provides superior pattern recognition capabilities.
Solution Approach 2:
The 3DCNN model acts as an intermediary between raw fMRI data and functional mapping results. The model processes the complex, noisy fMRI signals through multiple computational stages (convolutional layers, activation functions, pooling operations) to extract meaningful network information, serving as an intelligent mediator that enhances reliability.
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
The 3DCNN method allows for high-quality brain mapping in a shorter MRI time, improving accuracy and reducing patient discomfort, while maintaining or exceeding the performance of conventional methods.
Implementation Method 1
fMRI detects changes in the blood oxygen level dependent (BOLD) signal that reflect the neurovascular response to neural activity
Data Source
AI summary
A method for mapping brain function of a subject includes receiving a dataset of resting state fMRI (RS-fMRI) three dimensional (3D) image frames of the subject's brain, and inputting the 3D image frames to a deep learning artificial neural network. For each voxel of each 3D image frame and for each resting state network of a plurality of resting state networks, the deep learning artificial neural network calculates a probability that the voxel belongs to the resting state network. The deep learning artificial neural network is trained beforehand using a plurality of 3D image frames including previously defined resting state networks obtained from a plurality of calibration subjects. The method includes generating one or more functional map of the plurality of resting state networks of the subject's brain using the probabilities calculated by the artificial neural network.