Machine Learning Contrast Boost for Low-Contrast CT Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The use of contrast media in diagnostic imaging procedures poses health risks and economic challenges, with existing technological enablers being expensive, complex, or difficult to use, necessitating a need for alternative systems to reduce contrast agent usage.
Innovation Solution
A training system for a machine learning model that enhances image quality by processing high and low contrast imagery using a generative adversarial network to adjust parameters, allowing for reduced contrast media usage while maintaining diagnostic quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contrast media is administered to improve tissue visualization, then diagnostic quality is improved, but patient health risks and costs increase
Solution Approach 1:
The patent applies parameter changes by transforming the imaging approach from contrast-dependent to contrast-independent through machine learning processing. The system changes the parameter of image enhancement from chemical contrast agents to computational algorithms, thereby maintaining diagnostic quality while eliminating contrast media-related health risks
Solution Approach 2:
The patent substitutes the mechanical/chemical system of contrast media administration with a computational system. Instead of using physical contrast agents to enhance tissue visualization, the system uses machine learning models to process and enhance images, replacing the chemical mechanism with an algorithmic one
2Measurement precision
If contrast media is administered to improve tissue visualization, then diagnostic quality is improved, but cost increases
Solution Approach 1:
The patent employs cheap computational resources instead of expensive contrast agents. The machine learning models use inexpensive processing power to achieve image enhancement that previously required costly contrast media, effectively replacing expensive substances with affordable computational methods
3Quantity of substance
If technological enablers are used to reduce contrast usage, then contrast utilization is reduced, but device complexity and difficulty of use increase
Solution Approach 1:
The patent implements self-service by enabling the imaging system to automatically enhance images through machine learning without requiring manual intervention or complex operational procedures. The system serves itself by autonomously processing images to compensate for reduced contrast usage, eliminating the need for operators to manage complex reduction protocols
Data Source
AI summary
A training system for a target machine learning model for image enhancement and related methods. The system has a framework of two machine learning models (G1, G2) of the generative type, one such model (G1) being part of the target machine learning model. The training is based on a training data set including at least two types of training imagery, high image quality (IQ) imagery and low IQ imagery, the training input image being one of the high IQ type. The generative network (G1) processes the training input image of the high IQ type to produce a training output image having reduced IQ. The target machine learning model further produces, based on the training output image having the reduced IQ and the training input image of the high IQ type, a second training output image. The second generator network (G2) estimates, from the second training output image, an estimate of the training input image of the high IQ type. A training controller adjusts parameters of the machine learning model framework, based on a deviation between the estimate of the training input image of the high IQ type and the training input image of the high IQ type.