Low-Field MRI Image Enhancement Over Secure Metasurface Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution magnetic resonance imaging (MRI) devices are costly and not feasible for widespread use due to their high magnetic field strength and price, while lower-strength portable MRIs produce low-quality images unsuitable for medical diagnostics, necessitating a cost-effective method to enhance image resolution and ensure data privacy during transmission.
Innovation Solution
A portable MRI device captures low-resolution images, which are enhanced to high-resolution using a cycle generative adversarial network and annotated by a conditional GAN, with local training and federated learning, and secured through a reconfigurable intelligent surface for hardware-encrypted transmission in a private 5G network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high magnetic field strength (1.5-3.0T) is used to achieve high image quality, then diagnostic capability is improved, but device cost and size increase significantly
Solution Approach 1:
The patent segments the MRI system into two parts: a portable low-field MRI device for data acquisition and a separate processing system using deep learning models for image enhancement. This separation allows the imaging device to remain small and affordable while the computational resources are distributed to edge or cloud infrastructure.
Solution Approach 2:
The patent introduces deep learning models as an intermediary between the low-field MRI data and the final diagnostic images. These models act as a bridge that transforms low-quality images into high-quality diagnostic images without requiring high magnetic field strength.
2Device complexity
If portable low-field MRI is used to reduce cost and size, then device accessibility is improved, but image quality deteriorates to unsuitable levels for medical diagnostics
Solution Approach 1:
The patent applies deep learning models in advance to the low-resolution MRI images to predict and generate corresponding high-resolution images before diagnostic interpretation. This preliminary enhancement step ensures that portable devices can produce diagnostically useful images without requiring post-acquisition processing hardware upgrades.
Solution Approach 2:
The patent changes the parameter being optimized from magnetic field strength to computational processing. Instead of increasing field strength to improve image quality, the system maintains low field strength and compensates through algorithmic enhancement using trained neural networks.
3Productivity
If MRI data is transmitted over public networks for processing, then computational capability is improved, but data security and patient privacy are compromised
Solution Approach 1:
The patent segments the computational processing into local edge processing and selective cloud processing. Sensitive operations are performed locally on edge devices, while only non-sensitive model updates or aggregated data are transmitted to cloud infrastructure for federated learning.
Solution Approach 2:
The patent introduces federated learning as an intermediary mechanism that enables cloud-based model training without transmitting patient data to the cloud. The learning process occurs locally on edge devices, and only model parameter updates are shared, preserving data privacy while leveraging cloud computational resources.
Data Source
AI summary
The technology described herein is directed towards using a trained artificial intelligence (AI) model to generate high-resolution images from lower resolution magnetic resonance imaging (MRI) images captured by a lower magnetic field strength MRI device. For security and privacy, a reconfigurable intelligent surface can be used in the signal path to the trained model to thwart potential eavesdroppers. Also described is a trained AI annotator model that produces annotation data for annotating a generated high-resolution image. Local training using a cycle generative adversarial network, and based in part on federated learning, provides a highly-accurate low-resolution-to-high-resolution image generator model, while a conditional generative adversarial network provides a highly-accurate annotator model. A medical expert can thus analyze the highly-accurately generated high-resolution images with the benefit of annotation data to highlight any defects detected by the annotator model.