Parametrized Medical Image Conversion Function Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting medical images from one type to another, such as CT to MR, face inaccuracies due to reliance on unpaired data or lack of comparison with original images, leading to potential false information and reduced accuracy.
Innovation Solution
A system and method that generates an optimized parametrized conversion function by comparing converted images with original images, using penalties based on image pair similarities and discriminator errors to iteratively adapt and improve the conversion process, incorporating paired data and discriminator functions for enhanced accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If unpaired data is used for training the conversion function, then the method can work with available data without requiring precise matching, but the generated conversion function becomes less accurate and may introduce false information
Solution Approach 1:
The patent implements feedback by calculating a penalty function that compares the converted image with the paired ground truth image. This penalty is fed back to update the conversion function parameters, ensuring that even when trained on unpaired data, the function learns from the available paired comparisons to improve accuracy and reduce false information.
Solution Approach 2:
The patent performs preliminary action by first establishing paired image relationships to create a reference standard. This paired comparison framework is set up before the actual conversion process, allowing the system to evaluate and refine the conversion function against known correct mappings, thereby improving accuracy despite using unpaired training data.
2Device complexity
If no comparison is made between converted image and original image, then the training process is simpler, but the generated conversion function becomes less accurate
Solution Approach 1:
The patent introduces feedback through the penalty calculation that explicitly compares converted images with original paired images. This comparison mechanism provides directional guidance for optimizing the conversion function, ensuring higher accuracy while keeping the training process manageable through automated gradient-based optimization.
Solution Approach 2:
The patent replaces manual or complex mechanical comparison processes with automated computational methods. The penalty function uses algorithmic image comparison and gradient descent optimization to automatically adjust conversion parameters, substituting complex manual tuning with efficient computational optimization that maintains high accuracy.
3Measurement precision
If paired data is required for accurate conversion, then conversion accuracy improves, but the data requirements become more stringent and harder to obtain
Solution Approach 1:
The patent performs preliminary action by using available paired data to establish reference standards and penalty functions. This preliminary setup allows the system to leverage even limited paired data effectively, improving accuracy without requiring extensive paired datasets. The paired comparisons are prepared in advance to guide the conversion process.
Solution Approach 2:
The patent makes the training system universal by enabling it to work with both paired and unpaired data. The penalty function framework can accommodate mixed data types, allowing the conversion function to be trained on whatever data is available. This multi-functionality reduces data requirement stringency while maintaining conversion accuracy through the unified training approach.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In accordance with one or more embodiments herein, a system 100 for generating an optimized parametrized conversion function T for converting an original medical image of a first image type into a converted medical image of a second image type is provided. The system 100 comprises at least one processing unit configured to: obtain original medical images of the first and second image types; obtain an initial parametrized conversion function G to convert original medical images of the first image type into converted medical images of the second image type; calculate a first penalty P1 based on at least one comparison of a first original medical image of the second image type with a first converted medical image of the second image type, that has been generated by applying a first parametrized conversion function G1, which is based on the initial parametrized conversion function G, to a first original medical image of the first image type that forms an image pair with the first original medical image of the second image type and has thereby been determined to show the same part of the same patient; calculate a second penalty P2 based on at least one comparison of an original medical image of the first image type and a converted medical image of the second image type that has been generated by applying a second parametrized conversion function G2, which is based on the initial parametrized conversion function G, to the original medical image of the first image type, after converting the original medical image of the first image type and/or the converted medical image of the second image type into images of the same image type; and generate the optimized parametrized conversion function T based on the parameters of the initial parametrized conversion function G and at least said first and second penalties P1 and P2.