Neural Network Soft Tissue Estimation from DXA Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating the composition of soft tissues, such as muscle and fat, in subjects using radiation images lack accuracy and require multiple imaging techniques or complex data processing.
Innovation Solution
A device and method utilizing a trained neural network that processes simple radiation images or DXA scanning images, learning from images with different energy distributions and CT images to derive accurate composition information, including muscle mass, fat mass, and disease risk indicators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple imaging techniques (DXA, CT) are used to improve soft tissue composition estimation accuracy, then measurement precision improves, but device complexity and loss of time increase
Solution Approach 1:
The patent segments the imaging process by separating bone and soft tissue analysis. The neural network is trained to independently estimate soft tissue composition parameters from DXA images after bone information is extracted, allowing each tissue type to be analyzed with optimized methods rather than requiring full CT scanning for both.
Solution Approach 2:
The patent introduces a neural network as an intermediary that processes DXA image data to generate accurate soft tissue composition estimates. This intermediary transforms the limited DXA image information into detailed composition data, effectively bridging the gap between simple imaging and complex multi-modal imaging requirements.
2Measurement precision
If multiple imaging techniques (DXA, CT) are used to improve soft tissue composition estimation accuracy, then measurement precision improves, but loss of time increases
Solution Approach 1:
The patent performs preliminary bone information extraction from the DXA image before soft tissue composition estimation. This preliminary action separates the bone and soft tissue analysis pipelines, allowing the soft tissue estimation to proceed independently and rapidly without waiting for full CT processing.
Solution Approach 2:
The patent replaces the mechanical CT scanning process with a computational neural network approach. Instead of physically scanning the patient with CT equipment, the system uses a trained neural network to infer soft tissue composition from standard DXA images, dramatically reducing imaging time while maintaining accuracy.
3Ease of operation
If simple imaging is used to reduce device complexity and imaging time, then ease of operation improves, but measurement precision deteriorates
Solution Approach 1:
The patent changes the parameters extracted from simple DXA images by using a neural network to derive detailed composition parameters (fat mass, lean mass, bone mineral content) that go beyond what traditional DXA analysis provides. This transforms limited image data into comprehensive composition information through intelligent parameter extraction.
Solution Approach 2:
The neural network serves as an intermediary that enhances the information content of simple DXA images. It processes the basic image data and generates detailed soft tissue composition estimates, effectively upgrading the measurement capability without requiring more complex imaging equipment.
4Measurement precision
If neural network training with multiple data types is used to improve estimation accuracy, then measurement precision improves, but device complexity and loss of time increase
Solution Approach 1:
The patent performs preliminary processing of training data including CT images and DXA images to extract relevant features before neural network training. This preliminary action organizes the complex multi-source data into structured formats, simplifying the training process and reducing the computational burden during deployment.
Solution Approach 2:
The patent extracts only the essential soft tissue composition parameters from the complex multi-modal training data for neural network learning. By focusing on key parameters (fat mass, lean mass, bone mineral content) rather than processing all available data, the system reduces processing complexity while maintaining estimation accuracy.
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
Enables high-accuracy estimation of soft tissue composition and disease risk assessment by leveraging multiple imaging modalities and neural network processing, improving diagnostic precision.
Implementation Method 1
two radiation images acquired by imaging the subject with radiation having different energy distributions
Implementation Method 2
radiation incident on and transmitted through a human body, which is attenuated by a mass attenuation coefficient
Data Source
AI summary
At least one processor is provided, in which the processor functions as a trained neural network that derives an estimation result relating to a composition of a soft tissue of a subject from a simple radiation image acquired by simply imaging the subject, or a DXA scanning image acquired by imaging the subject by a DXA method. The trained neural network learns using, as teacher data, two radiation images acquired by imaging the subject with radiation having different energy distributions, the radiation image of the subject and a soft part image representing the soft tissue of the subject, or a composite two-dimensional image representing the subject derived by combining a three-dimensional CT image of the subject, and information relating to the composition of the soft tissue of the subject.