Reconstruction Network and Method for Reconstructing Cine Mri Images

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention presents a machine learning-based reconstruction network and methodology that enables fast, high-quality reconstruction of cine (dynamic) MRI images—particularly those used in heart imaging. The approach uses a variational network architecture composed of multiple cascaded modules, each acting as an update step, to process a stacked sequence of MRI frames. By focusing the reconstruction on a single, strategically-selected frame (processed by the network), it produces sharper, more reliable images with minimal delay (latency). The method is well-suited for interactive, real-time MRI imaging and can be adapted for both the reconstruction network and its training.

Use CasesContent extracted from patent full text and abstract with AI.

• Real-time cardiac MRI imaging for diagnosis of heart function and conditions.
• Interactive MRI-guided cardiac interventions where clinicians need immediate high-quality images to guide procedures.
• Any dynamic organ or motion imaging using cine MRI (lungs, joints, fetal movement, etc.).
• Remote or cloud-based MRI image reconstruction to off-load computational demands from scanner hardware.
• MRI research and development, enabling faster trials and experimental protocols involving dynamic imaging.

BenefitsContent extracted from patent full text and abstract with AI.

• Greatly reduced image reconstruction latency, supporting real-time medical decision making and interventions.
• Improved image quality from undersampled MRI data, allowing for faster data acquisition and reduced patient scan time.
• Enhanced clarity and reduced motion artifacts in the resulting images, as the network leverages information from a sequence of frames.
• Flexible system architecture that can balance between latency and image quality according to the needs of the application.
• Scalable and upgradable (software-based) solution—can be implemented via software or upgraded via cloud services on existing MRI hardware.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The invention describes a reconstruction network (12) for reconstructing cine MRI images, the reconstruction network (12) being a variational network designed for reconstructing images from cine MRI data and comprising an architecture with a cascade of cascade-modules (C1, C2, C3), wherein the input of the first cascade-module (C1, C2, C3) is an input-stack (S) of a plurality of N frames (K1, K2, K, K4) and the input of each following cascade-module (C1, C2, C3) is the input-stack (S) and an output-stack (P, P') of the preceding cascade-module (C1, C2), wherein the reconstruction network (12) comprises a selection-unit (U) designed to select one single frame (F) being processed by the cascade-modules (C1, C2, C3) that corresponds to the i-th frame of the input stack (S) for being the basis for the output dataset. The invention further describes a reconstruction method, a training method and a related magnetic resonance imaging system.