X-Ray Bone Density Estimation Using Synthetic DRR Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting bone density from X-ray images require a large amount of training data and have high processing loads, limiting their practical application in medical settings.
Innovation Solution
A non-transitory computer-readable storage medium that stores a program to generate a learning model capable of estimating bone density from X-ray images using a small number of cases by training with associated CT images, employing a Generative Adversarial Network (GAN) to produce Digital Reconstructed Radiographs (DRR) images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of training data is used to train the deep learning model for bone density prediction, then the prediction accuracy is improved, but the processing load and data collection requirements increase significantly
Solution Approach 1:
The patent performs preliminary actions by generating synthetic training data through simulated X-ray images and DRR images before actual model training. This pre-generation of training data reduces the need for collecting large amounts of real patient data, thereby improving data collection efficiency while maintaining prediction accuracy.
Solution Approach 2:
The patent creates copies of real X-ray images through synthetic generation and DRR (Digital Reconstructed Radiograph) techniques. These synthetic copies serve as training data, reducing the dependency on large volumes of actual patient data while preserving the diagnostic information needed for accurate bone density prediction.
2Measurement precision
If a large amount of training data is collected and processed, then the model training accuracy is improved, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary data preparation by generating synthetic training pairs (X-ray images and corresponding DRR images) before model training. This pre-processing step reduces the computational burden during actual training by providing ready-to-use, high-quality training data that requires minimal additional processing.
Solution Approach 2:
The patent uses a controlled amount of synthetic training data that is sufficient for achieving good model performance without requiring exhaustive data collection. By generating an appropriate quantity of synthetic samples rather than collecting all possible real data, the method achieves good accuracy while reducing processing time.
3Measurement precision
If complex deep learning models are used for bone density prediction, then the prediction accuracy is improved, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent introduces DRR (Digital Reconstructed Radiograph) images as an intermediary between real X-ray images and bone density measurements. The model learns to map X-ray images to DRR images, and then DRR images to bone density values. This intermediate representation simplifies the learning task and reduces computational complexity while maintaining prediction accuracy.
Solution Approach 2:
The patent segments the prediction task into two distinct stages: first mapping X-ray images to DRR images, then mapping DRR images to bone density values. This segmentation allows each stage to be optimized independently, reducing the overall computational complexity compared to a single direct mapping approach.
Data Source
AI summary
Provided is a program, etc. capable of acquiring information related to an amount of body tissue from an X-ray image with high accuracy using a small number of cases. A computer acquires training data including an X-ray image of a target site and information related to an amount of body tissue obtained from a CT (Computed Tomography) image of the target site. The computer generates a learning model configured to output information related to an amount of body tissue of a target site in an X-ray image when the X-ray image is input using acquired training data.