Synthetic Perfusion Mapping From Non-Contrast CT for Stroke Triage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computed tomography perfusion (CTP) imaging for brain hemodynamics requires intravenous contrast agents, which is costly, time-consuming, and poses health risks, making it suboptimal for rapid triage in acute ischemic stroke.
Innovation Solution
A deep learning framework, Model MAGIC, uses non-contrast CT scans to synthesize perfusion maps by leveraging physiology-informed multitask learning and a Physicians-in-the-Loop module to generate high-quality CTP maps without contrast agents, incorporating novel loss functions and a physician-guided adjustment mechanism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contrast agents are used for CTP imaging, then perfusion map quality and diagnostic accuracy are improved, but healthcare costs and patient health risks increase
Solution Approach 1:
The patent extracts and removes the contrast agent component from the CTP imaging process. The deep learning model learns to predict perfusion parameters directly from non-contrast CT images, eliminating the need for contrast agents while maintaining diagnostic accuracy. This is achieved through training the model on paired non-contrast CT and contrast-enhanced CTP images to learn the mapping relationship between anatomical structures and perfusion characteristics.
Solution Approach 2:
The patent creates synthetic contrast-enhanced CTP images by copying and transforming non-contrast CT images through the deep learning model. The model generates predicted perfusion maps that replicate the diagnostic information normally obtained from contrast agents, effectively creating a virtual copy of the contrast-enhanced imaging process without using actual contrast materials.
2Measurement precision
If contrast agents are used for CTP imaging, then perfusion map quality is improved, but imaging time and processing complexity increase
Solution Approach 1:
The patent performs preliminary action by pre-training the deep learning model on large datasets of paired non-contrast CT and contrast-enhanced CTP images. This preliminary training phase enables the model to learn complex perfusion patterns and relationships, so that during actual clinical use, perfusion maps can be generated instantly from non-contrast images without requiring the time-consuming contrast agent administration and multi-phase imaging protocols.
3Productivity
If non-contrast CT scans are used instead of CTP, then healthcare costs and imaging time are reduced, but perfusion information and diagnostic accuracy are lost
Solution Approach 1:
The patent introduces a deep learning model as an intermediary between non-contrast CT images and perfusion information. The model acts as a translator that converts anatomical information from non-contrast CT scans into functional perfusion parameters, bridging the information gap without requiring actual contrast-enhanced imaging. This intermediary enables the extraction of perfusion information that would otherwise be unavailable from non-contrast images alone.
Data Source
AI summary
Various examples are provided related to predicting perfusion images from non-contrast scans. In one example, a method for predicting perfusion images includes generating perfusion maps of an organ of a subject from non-contrast computed tomography (NCCT) slices of the organ; processing the perfusion maps based upon weights determined by a Physicians-in-the-Loop (PILO) module; and generating synthetic computed tomography perfusion (CTP) maps from the processed perfusion maps, the synthetic CTP maps generated by deep learning-based multimodal image translation. In another example, a system includes at least one computing device that can generate prefusion maps of an organ from NCCT slices; process the perfusion maps based upon weights determined by a PILO module; and generate synthetic CTP maps from the processed perfusion maps using deep learning-based multimodal image translation. The CTP maps can be rendered for display to a user.