Radiology Contrast Transfer for Washout-Resistant Tomograms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radiological imaging techniques using contrast agents face challenges with rapid washout of contrast agents during long scan times, leading to decreased contrast and poorer image quality in later-acquired tomograms.
Innovation Solution
A method using a prediction model trained on a dataset to transfer contrast information from high-contrast tomograms to low-contrast tomograms, generating artificial, contrast-enhanced tomograms by rearranging the sequence of layer scanning and utilizing machine learning to enhance image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contrast agents are used to enhance vascular visualization in radiological imaging, then image contrast and diagnostic accuracy are improved, but contrast agent washout occurs rapidly during long scan times causing contrast to decrease continuously
Solution Approach 1:
The patent applies preliminary action by using machine learning models to predict and generate contrast-enhanced tomograms before the actual scan completes. The system pre-processes early-stage tomograms with high contrast information to create enhanced versions that can be used immediately, effectively preparing the image data in advance to compensate for the upcoming contrast washout during the scanning process.
Solution Approach 2:
The patent uses copying by creating artificial tomograms that replicate the appearance of high-contrast images. The machine learning model generates synthetic tomograms that copy the contrast characteristics from early-stage scans, producing virtual images that mimic the desired high-contrast appearance even when the actual contrast agent concentration has decreased in later scan stages.
2Productivity
If tomograms are scanned layer by layer to generate multiple tomograms, then complete volumetric imaging is achieved, but contrast enhancement decreases continuously from tomogram to tomogram due to washout
Solution Approach 1:
The system performs preliminary contrast enhancement on early-acquired tomograms before processing later tomograms. By applying machine learning-based contrast transfer in advance, the system ensures that all tomograms in the volume have consistent contrast levels, preventing the degradation that would otherwise occur as contrast agent washout progresses through the sequential scanning process.
Solution Approach 2:
The patent implements feedback by using machine learning models that continuously monitor and adjust contrast enhancement based on the scan progress. The model receives feedback about the actual contrast levels in different tomograms and dynamically adjusts the enhancement parameters to maintain uniform contrast across the entire volumetric dataset, compensating for the washout effect.
3Area of stationary object
If scan time is extended to capture complete anatomical structures, then imaging coverage is improved, but contrast agent concentration decreases leading to poorer image quality in later regions
Solution Approach 1:
The patent creates artificial tomograms that copy the high-contrast characteristics from early scan regions and apply them to later regions. The machine learning model generates synthetic images that replicate the desired image quality, effectively copying the contrast properties from regions where the contrast agent is still present to regions where it has washed out, thereby maintaining consistent image quality across the entire imaging coverage.
Solution Approach 2:
The machine learning model acts as an intermediary between the raw scan data and the final images. It processes the tomograms through neural networks that transfer contrast information from high-contrast early-stage images to later-stage images, serving as a mediating transformation that preserves image quality throughout the extended scan duration and comprehensive anatomical coverage.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to the technical field of producing artificial contrast-enhanced radiological images by way of machine learning methods.