Synthetic Medical Image Generation via Iterative Parameter Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Synthetic medical images generated using machine learning models are often blurred, making it difficult to recognize structures compared to measured medical images.
Innovation Solution
A computer-implemented method for training a machine learning model to generate synthetic images by providing training data that includes input and target data for each examination object, determining target image features, and modifying model parameters to reduce deviations between synthetic and target images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a machine learning model is used to generate synthetic medical images, then the productivity of image generation is improved, but the manufacturing precision (image quality and structure recognizability) deteriorates due to blurring
Solution Approach 1:
The patent implements a feedback mechanism where the generated synthetic image is compared with the target measured image, and the model parameters are iteratively adjusted based on the deviation between them. This closed-loop feedback process continues until the synthetic image quality meets the desired precision threshold, thereby resolving the contradiction between fast generation and high quality.
Solution Approach 2:
The patent performs preliminary actions by pre-processing both the input medical image and target image to extract relevant features and characteristics before the main synthesis process. This preliminary preparation enables the model to focus on critical structural elements during generation, improving image quality without significantly increasing computation time.
2Manufacturing precision
If the machine learning model parameters are adjusted to improve image quality, then the manufacturing precision is improved, but the device complexity increases
Solution Approach 1:
The patent applies local quality by adjusting specific model parameters that have the most significant impact on image quality rather than uniformly complexifying the entire model. The iterative optimization process identifies and refines only the critical parameters needed to reduce deviations between synthetic and target images, maintaining model simplicity while improving output quality.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Systems, methods, and computer programs disclosed herein relate to training a machine learning model and using the trained machine learning model to generate synthetic images, preferably synthetic medical images.