Deep Learning Network Converts Non-Contrast to Contrast Medical Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Patients with reduced renal function, particularly the elderly, face challenges in medical imaging as contrast media used in CT and MRI scans are nephrotoxic, necessitating the conversion of non-contrast images into contrast images without administering harmful substances.
Innovation Solution
An image conversion method and device utilizing a deep learning network trained with contrast and non-contrast learning images to generate contrast images from non-contrast images, specifically employing Maximum Intensity Projection (MIP) images to enhance learning performance and minimize error.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contrast medium is administered to improve tissue contrast and lesion distinguishability, then diagnostic clarity is improved, but nephrotoxicity increases making the method unsuitable for patients with reduced renal function
Solution Approach 1:
The patent creates a synthetic copy of contrast-enhanced images by training a deep learning model on paired datasets of non-contrast and contrast-enhanced images. The model learns to generate contrast-like images from non-contrast inputs, effectively copying the visual characteristics of contrast-enhanced images without using actual contrast media.
Solution Approach 2:
The patent replaces the physical-chemical mechanism of contrast media administration with a computational algorithm. Instead of using iodinated or gadolinium-based contrast agents that enhance X-ray or magnetic signal differentiation, the system uses a trained neural network to computationally enhance tissue contrast from non-contrast images.
2Manufacturing precision
If deep learning network is trained with Maximum Intensity Projection (MIP) images to improve learning performance, then image conversion accuracy is improved, but training data processing complexity increases
Solution Approach 1:
The patent performs Maximum Intensity Projection on contrast-enhanced images before training the deep learning model. By pre-processing the training data to create MIP images that emphasize contrast-enhanced structures, the model learns more effective features for generating accurate contrast-like images from non-contrast inputs.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and a device for converting a non-contrast image into a contrast image are disclosed. The image conversion device converts the non-contrast image into the contrast image by using a deep learning network trained with learning data including one or more contrast learning images and one or more non-contrast learning images.