Synthetic CTA Imaging With Multi-Scale Discrimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical image modality conversion models, such as MedGAN, fail to effectively highlight important zones in medical images due to the lack of consideration for windowing operations and regional differences, and iodine-based contrast agents pose risks for subjects with allergies or renal issues.
Innovation Solution
A method for synthesizing computerized angiography imaging using a multi-scale discriminator that includes a generator and a multi-scale discriminator with global and local discriminators, trained on normalized CT and CTA images to enhance the accuracy of discrimination and compatibility with existing equipment, eliminating the need for iodine contrast agents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional medical image conversion models (e.g., MedGAN) are used, then image conversion capability is achieved, but the ability to highlight important zones is insufficient due to lack of windowing operation consideration
Solution Approach 1:
The discriminator is segmented into multiple independent discriminator groups, each corresponding to a specific windowing operation. Each group contains global and local discriminators that independently evaluate different aspects of the generated images under different windowing conditions, enabling precise discrimination without requiring a monolithic complex model
Solution Approach 2:
The patent introduces a multi-scale discrimination dimension by incorporating windowing operations at different scales. The discriminator evaluates images not only at the original scale but also at scaled-down versions through windowing, adding a new dimension to the discrimination process that enhances the ability to detect important zones
2Illumination intensity
If iodine-based contrast agents are used for CT angiography, then tissue contrast enhancement is achieved, but safety risks increase for subjects with allergies or renal insufficiency
Solution Approach 1:
The patent creates synthetic CTA images that copy the appearance and characteristics of real CTA images without using actual contrast agents. The generator network learns from real CTA images and synthesizes new images that replicate the contrast-enhanced appearance, providing the same diagnostic information without the harmful effects of iodine-based contrast agents
Solution Approach 2:
The patent replaces the physical contrast agent mechanism with a computational mechanism. Instead of injecting contrast agents to enhance tissue visibility, the system uses deep learning models to computationally generate enhanced images from non-contrast CT scans, substituting a physical chemical process with an information processing approach
3Manufacturing precision
If synthetic CTA images are generated without multi-scale discrimination, then generation speed is maintained, but the ability to match real CTA image characteristics is insufficient
Solution Approach 1:
The patent performs preliminary actions during training by pre-computing and storing windowed versions of real CTA images along with their corresponding non-contrast CT images. This allows the generator to learn the relationship between different windowing scales in advance, enabling fast inference without requiring real-time multi-scale processing during image generation
Data Source
AI summary
The present invention discloses a method for synthesizing computerized angiographic imaging based on multi-scale discrimination, which includes generating a normalized training dataset and a normalized validation dataset; constructing a generator and a multi-scale discriminator; training the generator and the multi-scale discriminator based on the normalized training dataset; normalizing the non-contrast CT image to be processed and inputting it into the trained generator G to output a normalized synthetic CTA image; and restoring the normalized synthetic CTA image to its original pixel value range to obtain the synthetic CTA image. The present invention employs a multi-scale discriminator to perform multi-scale discrimination on the output of the generator, enabling the synthesized CTA images to highlight the target images specified by the windowing operation parameters and designated regions, thereby enhancing the accuracy of discrimination.