Pseudo 2D Breast Image Generation With Compression-Matched Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image generation models trained with a combination of tomosynthesis imaging and normal two-dimensional images captured in different compression states result in decreased accuracy of pseudo two-dimensional image generation.
Innovation Solution
An image generation apparatus and method that trains an image generation model using a combination of projection images or tomographic images and normal two-dimensional images captured in the same compression state, and further categorizes training data by factors such as person under examination and imaging settings to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the image generation model is trained using a combination of tomosynthesis imaging and normal two-dimensional images captured in different compression states, then the training process can be simplified and data requirements are reduced, but the generation accuracy of pseudo two-dimensional images decreases
Solution Approach 1:
The patent applies parameter changes by filtering training data based on compression state parameters. The system identifies and selects only those image combinations where the compression state parameter matches between tomosynthesis and mammography images, thereby improving generation accuracy while maintaining manageable training complexity through parameter-based data selection
2Manufacturing precision
If the image generation model is trained using combinations of images from the same breast in the same compression state, then the generation accuracy of pseudo two-dimensional images is improved, but the data selection process becomes more complex and requires additional filtering
Solution Approach 1:
The patent applies local quality by applying different selection criteria to different aspects of the training data. Specifically, it filters data based on local characteristics such as compression state, breast region, and image type, allowing the system to maintain high generation accuracy while managing data selection complexity through targeted filtering rather than comprehensive reprocessing
Solution Approach 2:
The patent uses an intermediary approach by introducing a compression state identification module that acts as a mediator between the raw image data and the training process. This intermediary component standardizes the data selection process by first determining compression states and then using this information to filter appropriate image combinations, thereby reducing overall process complexity
3Productivity
If training data is not filtered by compression state and breast region, then the training process is faster and requires less computational resources, but the training data contains inconsistent information that reduces model performance
Solution Approach 1:
The patent applies preliminary action by performing compression state determination and data filtering before the actual training process begins. The system pre-processes the training data to identify and select only suitable image combinations based on matching compression states and breast regions, ensuring high model performance while maintaining efficient training speeds through pre-filtered data
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enhances the generation accuracy of pseudo two-dimensional images by aligning training conditions with actual imaging states, leading to more precise image reproduction.
Implementation Method 1
a radiation source that emits radiation in an imaging direction, a detector that detects the radiation transmitted through the breast
Implementation Method 2
in a state in which a breast is compressed by a compression member
Data Source
AI summary
An image processing apparatus as an image generation apparatus includes an image generation model that has been trained in advance using a plurality of combinations of a plurality of projection images or a plurality of tomographic images and a normal two-dimensional image captured by irradiating, with radiation, a breast in a state of being compressed by a compression member during tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images.