Cross-Modal Elastography Mapping for Liver Fibrosis Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current liver fibrosis diagnosis methods, such as ultrasound and magnetic resonance shear wave elastography, face challenges in comparability due to differences in excitation frequency and post-processing algorithms, leading to inconsistent and costly follow-up examinations.
Innovation Solution
A method and system that uses deep learning algorithms, specifically generative adversarial networks (GAN) and convolutional neural networks (CNN), to predict corresponding stiffness values and confidence levels between ultrasound and magnetic resonance elastography, enabling a smart dashboard for consistent liver fibrosis monitoring across imaging modalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If ultrasound shear wave elastography (UE) is used for liver fibrosis monitoring, then cost is reduced and availability is improved, but measurement precision and comparability with MRE deteriorate
Solution Approach 1:
The patent introduces a deep learning-based translation model as an intermediary that converts UE stiffness maps into MRE-equivalent stiffness maps. This mediator enables comparability between the two modalities by translating UE measurements into the MRE reference framework, allowing clinics to use the more accessible UE while maintaining measurement comparability through AI-based translation.
Solution Approach 2:
The patent transforms the measurement parameters by using deep learning to map UE stiffness values and spatial distributions into the MRE parameter space. The translation model learns the parameter transformation between different imaging modalities, enabling UE (with its different excitation frequency and bandwidth characteristics) to provide measurements comparable to MRE.
2Measurement precision
If MRE is used for liver fibrosis diagnosis, then measurement precision and reliability are improved, but device complexity and patient discomfort increase
Solution Approach 1:
The patent creates a virtual copy of the MRE measurement process by training a deep learning model on paired UE-MRE datasets. The model learns to generate MRE-equivalent stiffness maps from UE inputs, effectively copying the MRE measurement outcome without requiring the complex MRE hardware and protocol, thus maintaining reliability while reducing device complexity requirements.
3Adaptability or versatility
If different imaging modalities (UE and MRE) are used for follow-up examinations, then adaptability and patient convenience are improved, but measurement consistency deteriorates
Solution Approach 1:
The deep learning translation model serves as a mediator that standardizes measurements from different imaging modalities into a common MRE-equivalent framework. This allows clinics to flexibly choose UE or MRE for different patient visits while the translation model ensures measurement consistency by converting all inputs to the MRE reference space, maintaining longitudinal comparability despite modality switching.
4Measurement precision
If liver biopsy is performed for fibrosis diagnosis, then measurement precision is improved, but patient harm and procedural complexity increase
Solution Approach 1:
The patent replaces the mechanical invasive biopsy procedure with non-invasive elastography imaging (UE or MRE) combined with deep learning translation. This substitution eliminates the physical harm and procedural complexity of needle insertion while maintaining diagnostic accuracy through AI-enhanced non-invasive stiffness measurement and fibrosis staging.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate and non-invasive liver fibrosis assessment by bridging the gap between ultrasound and magnetic resonance elastography, reducing costs and patient discomfort, while providing a user-friendly dashboard for monitoring stiffness changes over time.
Implementation Method 1
A continuous acoustic vibration at low frequency (∼60 Hz for liver) is transmitted into the region of interest (e.g., abdomen) of a subject using the passive driver, producing harmonic waves with a narrow frequency bandwidth
Implementation Method 2
UE methods, which differ from MRE methods, use high-intensity short-duration ultrasound 'push' pulses to generate a shear wave with a broad frequency spectrum
Implementation Method 3
Ultrasound imaging vendors use time-of-flight methods to reconstruct a shear wave speed, similar to the group velocity
Data Source
AI summary
A method and system (100) for augmented interpretation of shear wave elastography between first and second imaging modalities comprises performing an elastography measurement via a second imaging modality (20), different from a first imaging modality (10), to obtain at least one second imaging modality elastography value (32, 60) of a region of interest (33). At least one corresponding first imaging modality elastography value (36, 38, 62) is predicted based on the obtained second imaging modality elastography value. A graphical user interface or smart report dashboard (50) is generated that shows (i) a fibrosis level (521) of the region of interest, wherein the fibrosis level is determined as a function of (i)(a) the at least one second imaging modality elastography value (32) and/or (i)(b) the predicted at least one corresponding first imaging modality elastography value (36, 38).