MRI Scan Prediction for Dynamic Contrast Liver Imaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing radiological examinations using dynamic contrast-enhanced MRI for focal liver lesions require prolonged patient immobilization due to the need for multiple scans over time, leading to discomfort and potential motion artifacts.

Innovation Solution

A machine learning model is trained to predict subsequent MRI scans based on previous scans, learning the dynamics of contrast agent distribution, allowing for calculated scans rather than real-time acquisitions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple MRI scans are acquired over time to monitor contrast agent distribution, then diagnostic information quality is improved, but examination time increases and patient comfort deteriorates

Engineering Contradiction:
Improvediagnostic information qualityVSAvoidexamination time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent uses a machine learning model to generate synthetic MRI scans that copy the appearance and diagnostic information of real scans. The model learns from a training set of real MRI scans and generates synthetic images that mimic the contrast agent distribution patterns, thereby providing diagnostic information without requiring actual prolonged scanning.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary training of the machine learning model using a training set of MRI scans before the actual examination. This preliminary action enables the model to predict contrast agent distribution patterns, allowing the system to generate synthetic scans quickly during the examination without requiring prolonged real-time scanning.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If multiple MRI scans are acquired over time to monitor contrast agent distribution, then diagnostic information quality is improved, but patient comfort deteriorates due to prolonged immobilization

Engineering Contradiction:
Improvediagnostic information qualityVSAvoidpatient comfort
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent generates synthetic MRI scans that replicate the diagnostic information of real scans, eliminating the need for patients to remain immobilized for extended periods. The machine learning model creates copies of the required diagnostic images through computation rather than physical scanning.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The model is trained in advance on a comprehensive training set, enabling it to predict contrast enhancement patterns without requiring patients to undergo prolonged scanning sessions. This preliminary preparation transfers the time burden from the examination phase to the training phase.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If multiple MRI scans are acquired over time to monitor contrast agent distribution, then diagnostic information quality is improved, but motion artifacts increase

Engineering Contradiction:
Improvediagnostic information qualityVSAvoidmotion artifacts
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent generates synthetic MRI scans through machine learning computation rather than physical scanning. Since these images are computationally generated rather than acquired from moving patients, they are free from motion artifacts that plagues traditional prolonged MRI examinations.

Inventive Principle:
Principle #26Copying

4Loss of time

If a machine learning model is trained to predict subsequent MRI scans, then examination time is reduced, but model complexity increases

Engineering Contradiction:
Improveexamination timeVSAvoidmodel complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent employs a machine learning model that learns to copy the patterns of contrast agent distribution from a training set of MRI scans. This copying capability allows the model to generate synthetic scans that replicate real scan outcomes, reducing examination time while managing complexity through pattern recognition rather than complex physical scanning protocols.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4483197B1Prediction of a representation of an object to be examined in a state of a sequence of states
Publication Date: 2025.11.05 BAYER AG
  • EP4483197B1 patent drawingFigure 1(a)~2
  • EP4483197B1 patent drawingFigure 3(A)~3(D)
  • EP4483197B1 patent drawingFigure 4(A)~4(D)

AI summary

The present invention relates to the technical field of radiology, and in particular to assisting radiologists in radiological examinations using artificial intelligence methods. The present invention also relates to training a machine learning model and using the trained model to predict representations of an examination area in one or more states of a sequence of states in a radiological examination.