Simulating Higher Contrast Agent Doses in Medical Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging techniques using reduced-dose contrast agents often result in inadequate contrast enhancement, particularly for low-grade tumors, leading to difficulties in distinguishing targets from nearby features, which can lead to false positives/negatives, reduced therapy effectiveness, and tissue damage in diagnostic and surgical applications.
Innovation Solution
A machine learning model is trained to simulate higher doses of contrast agents by optimizing the capability to increase contrast from a source dose to a target dose, using a neural network that can generate simulation images mimicking administration of a contrast agent at a higher dose than the administered dose, allowing for enhanced contrast enhancement without the risks associated with full-dose administration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a reduced-dose of contrast agent is administered to minimize risks, then patient safety is improved, but contrast enhancement becomes inadequate making it difficult to distinguish targets from nearby features
Solution Approach 1:
A machine learning model acts as an intermediary between the reduced-dose images and the desired full-dose appearance. The model is trained on pairs of reduced-dose and full-dose images to learn the transformation, then applies this learned transformation to generate enhanced contrast images from new reduced-dose inputs, effectively mediating the gap between low and high contrast states
Solution Approach 2:
The system creates a simulated copy of what the images would look like with full-dose contrast agent administration. By training the machine learning model on full-dose images as ground truth, it generates synthetic full-dose images from reduced-dose inputs, copying the appearance characteristics without requiring actual full-dose exposure
2Measurement precision
If a full-dose of contrast agent is administered to achieve adequate contrast enhancement, then target visibility is improved, but patient exposure to contrast agent increases creating potential health risks
Solution Approach 1:
The system creates a simulated copy of what the images would look like with full-dose contrast agent administration. By training the machine learning model on full-dose images as ground truth, it generates synthetic full-dose images from reduced-dose inputs, copying the appearance characteristics without requiring actual full-dose exposure
Solution Approach 2:
The system changes the effective contrast parameter by applying a learned transformation that enhances the contrast in reduced-dose images to match full-dose levels. This allows achieving full-dose contrast enhancement效果 through post-processing parameter adjustment rather than increasing the actual contrast agent dose
3Measurement precision
If contrast enhancement is increased to improve target distinction, then diagnostic accuracy is improved, but the complexity of the imaging system increases requiring machine learning models
Solution Approach 1:
A machine learning model acts as an intermediary between the reduced-dose images and the desired full-dose appearance. The model is trained on pairs of reduced-dose and full-dose images to learn the transformation, then applies this learned transformation to generate enhanced contrast images from new reduced-dose inputs, effectively mediating the gap between low and high contrast states
Data Source
AI summary
A solution is proposed relating to medical imaging applications. Particularly, a method (600) for imaging a body-part of a patient comprises simulating (624-630) corresponding operative simulation images from an operative baseline image and operative administration images, which operative administration images have been acquired with administration of a contrast agent at an operative administration-dose: the operative simulation images mimic administration of the contrast agent at a higher dose. For this purpose, a machine learning model (420) is used that has been trained to optimize a capability thereof to mimic a corresponding increase of the contrast agent from a sample source-dose to a sample target-dose: the sample source-dose is different from the operative administration-dose. Corresponding computer program (500) and computer program product for implementing the imaging method (600) are proposed. Moreover, a computing system (115) for performing the imaging method (600) and an imaging system (105) comprising the computing system (115) and a scanner (110) are proposed. A medical method based on the same imaging method (600) is further proposed.