Ultrasonic Learning Model Inverse Coordinate Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of machine learning models used in ultrasonic diagnostic systems decreases due to information loss during coordinate transformation of medical images, making it difficult to detect clinically meaningful structures accurately.
Innovation Solution
A learning model is developed that uses a pair of ultrasonic image data before and after coordinate transformation, where the correct answer data is inversely transformed to maintain the original image information, allowing the model to learn from the intermediate processed image data, thereby improving learning accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If coordinate transformation is performed on ultrasonic image data to generate display images, then the images become easier to view and interpret, but information is lost and learning accuracy decreases
Solution Approach 1:
The patent applies preliminary action by performing inverse coordinate transformation on the correct answer data before training the learning model. This prepares the training data in the same coordinate system as the input image data, ensuring that the model learns to map from the original coordinate system directly to the transformed coordinate system, thereby preserving information while maintaining ease of interpretation.
Solution Approach 2:
The patent introduces an intermediary process of inverse transformation that bridges the gap between the transformed correct answer data and the original coordinate system image data. This intermediary step ensures that the learning model receives consistent coordinate system data, preventing information loss while maintaining the benefits of coordinate transformation for display.
2Measurement precision
If coordinate transformation is applied to ultrasonic image data, then the diagnostic display is improved, but the learning model accuracy deteriorates
Solution Approach 1:
The patent applies inversion by reversing the coordinate transformation process. Instead of transforming the input image data to match the correct answer coordinate system, the patent inversely transforms the correct answer data to match the input image coordinate system. This ensures that the learning model is trained on data in the same coordinate system, maintaining both diagnostic display accuracy and learning model accuracy.
3Device complexity
If transformed image data is used for machine learning training, then the processing workflow is simplified, but the detection precision of clinically meaningful structures decreases
Solution Approach 1:
The patent applies preliminary action by pre-processing the correct answer data through inverse coordinate transformation before training. This ensures that the training data is in the same coordinate system as the input images from the beginning, allowing the learning model to learn accurate mappings without requiring complex post-processing or coordinate system conversions during the learning process.
Data Source
AI summary
A non-transitory storage medium storing a computer-readable diagnostic program that causes a computer to execute outputting that is outputting a first inference result from third ultrasonic image data before processing including coordinate transformation based on a reception signal for image generation received by an ultrasonic probe by using a learning model. The learning model is machine-learned using learning data formed with a pair of: first ultrasonic image data based on a reception signal for image generation received by an ultrasonic probe; and second correct answer data obtained by performing inverse transformation of coordinate transformation on first correct answer data for second ultrasonic image data obtained by performing processing including coordinate transformation on the first ultrasonic image data.