Artificial Neural Network Perfusion Image Generation Without Contrast Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating perfusion images in medical diagnostic imaging require contrast agents, which can cause adverse effects and substantial procedural effort, and involve complex processes involving multiple imaging scans and subtraction calculations, leading to potential artifacts.
Innovation Solution
An apparatus and method using an artificial neural network trained on non-contrast medical diagnostic images to generate perfusion images without the need for contrast agents, reducing the number of imaging scans and processing steps, and leveraging U-net type deep learning networks for effective image segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contrast agents are used to generate perfusion images, then perfusion information can be obtained, but adverse effects occur and procedural effort increases
Solution Approach 1:
The invention extracts and removes the contrast agent administration step from the perfusion imaging process. By using non-contrast MRI sequences (T1, T2, DWI) combined with artificial neural network processing, the method obtains perfusion information without introducing contrast agents into the patient's body, thereby eliminating allergic reactions and deposition risks while maintaining diagnostic accuracy
Solution Approach 2:
The invention replaces the physical-chemical mechanism of contrast agent enhancement with a computational mechanism. Instead of relying on contrast agents to alter tissue signal properties, the system uses deep learning algorithms to process and transform non-contrast MRI data into perfusion images, substituting mechanical/chemical processes with information processing
2Measurement precision
If multiple imaging scans and subtraction calculations are performed, then perfusion images can be generated, but device complexity and processing steps increase
Solution Approach 1:
The invention merges multiple MRI sequence acquisitions (T1-weighted, T2-weighted, diffusion-weighted images) into a unified deep learning processing pipeline. The artificial neural network simultaneously processes all these sequences to generate perfusion parameters, eliminating the need for separate contrast-enhanced scans and manual subtraction operations, thereby reducing overall procedural complexity
Solution Approach 2:
The invention uses artificial neural networks to learn the mapping from non-contrast MRI sequences to contrast-enhanced perfusion images. The network is trained on paired data (non-contrast inputs and corresponding contrast-enhanced outputs) and then copies this learned transformation to generate perfusion images from new non-contrast scans, replacing the need for actual contrast administration and subtraction calculations
3Measurement precision
If contrast-enhanced MRI scans are acquired, then perfusion dynamics can be visualized, but patient stress and resource consumption increase
Solution Approach 1:
The invention extracts perfusion dynamics information from non-contrast MRI sequences through deep learning, removing the need for contrast agent administration. This eliminates the metabolic burden and stress associated with contrast agents while preserving the ability to visualize and quantify perfusion dynamics over time
Solution Approach 2:
The invention enables non-contrast MRI sequences to serve the dual purpose of both anatomical imaging and functional perfusion assessment. By using the inherently acquired non-contrast data combined with AI processing, the system makes the imaging protocol self-sufficient, eliminating the need for additional contrast-enhanced scans and reducing overall resource consumption
Data Source
AI summary
The invention provides an apparatus and a method for generating a perfusion image, as well as a method for training an artificial neural network for use therein. The method comprises at least steps of: receiving (S100) at least one non-contrast medical diagnostic image. NCMDI (1-i), acquired from organic tissue: generating (S200), using an artificial neural network. ANN (2), trained and configured to receive input data (10) based on at least one of the received at least one non-contrast medical diagnostic image, NCMDI (1-i), based on the input data (10), at least a perfusion image (3) for the organic tissue shown in the at least one non-contrast medical diagnostic image, NCMDI (1-i); and outputting (S300) at least the generated perfusion image (3).