Synthetic Contrast-Enhanced CT via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a lack of liver-specific intracellular contrast agents authorized for use in computed tomography (CT) imaging, limiting the ability to enhance visualization of liver tissues in CT scans.
Innovation Solution
A method is developed to train a machine learning model using CT data from examinations with different doses of MR contrast agents, allowing the generation of synthetic contrast-enhanced CT images that mimic the appearance of higher-dose contrast agent administration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MR contrast agents are used in CT imaging, then contrast enhancement is achieved, but the lack of liver-specific agents limits the ability to enhance visualization of liver tissues
Solution Approach 1:
The patent uses machine learning models to generate synthetic CT images that replicate the appearance of high-dose contrast agent administration. The model learns from training data containing CT images with different contrast agent doses and generates synthetic images that mimic the enhancement pattern of higher doses, effectively copying the visual effect without requiring actual high-dose administration.
Solution Approach 2:
The patent changes the parameter of contrast agent dose by using machine learning to simulate the effect of higher doses from lower-dose images. The training process involves images with different contrast agent concentrations, and the model learns to transform low-dose images into synthetic high-dose images, thereby changing the effective contrast enhancement level without physically changing the administered dose.
2Measurement precision
If higher doses of contrast agents are administered, then contrast enhancement is improved, but the need for higher doses increases the risk of side effects and cost
Solution Approach 1:
The patent creates synthetic CT images that copy the visual appearance of high-dose contrast enhancement using machine learning. The model takes low-dose CT images and generates synthetic versions that replicate the enhancement patterns, texture, and contrast characteristics of high-dose images, thereby achieving improved visualization without actually administering high doses.
Solution Approach 2:
The machine learning model acts as an intermediary between the actual contrast agent administration and the desired high-contrast images. Instead of directly administering high-dose contrast agents, the system uses the ML model to translate low-dose images into synthetic high-contrast images, mediating the relationship between minimal contrast agent use and optimal visualization quality.
3Measurement precision
If machine learning models are trained using CT data from multiple contrast agent doses, then synthetic high-quality images are generated, but the complexity of data collection and model training increases
Solution Approach 1:
The patent creates a universal training framework that can handle multiple contrast agent doses and image types within a single model. The machine learning model is designed to process CT images from various contrast phases and generate synthetic images across different enhancement patterns, making the system versatile for multiple imaging scenarios without requiring separate specialized models for each case type.
Data Source
AI summary
The present invention relates to systems, methods and computer programs for training and using a machine learning model to generate synthetic contrast-enhanced computed tomography images with the help of magnetic resonance contrast agents.