Synthetic Mammogram Image Generation via Frequency Component Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing maximum intensity projection (MIP) method for generating synthetic mammogram images can result in deteriorated image quality due to the loss of pixel value information between pixels, as it uses the maximum pixel value from three-dimensional data, which may include structures not present on the same cross-section, leading to noise influence and reduced contrast.
Innovation Solution
An image processing device and method that generates a second synthetic mammogram image by synthesizing high-frequency components from a first synthetic mammogram image with high-frequency components and low-frequency components from projection images captured at different angles, including the normal direction, to improve image quality by retaining more information and reducing noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If the maximum intensity projection (MIP) method is used to generate synthetic mammogram images, then noise resistance is improved and images with small contrast can be distinctly depicted, but image quality deteriorates due to loss of pixel value information between pixels
Solution Approach 1:
The patent segments the image processing into two distinct stages: first generating a synthetic mammogram image using the MIP method to ensure noise resistance, then performing frequency separation to extract high-frequency and low-frequency components. This segmentation allows each stage to optimize for its specific purpose while avoiding the degradation of overall image quality.
Solution Approach 2:
The patent changes the processing parameters by applying different spatial frequency thresholds to separate image components. By transforming the image into frequency domain and applying specific frequency cutoffs, the system preserves both the noise-resistant properties of MIP and the detailed structural information lost in conventional MIP processing.
2Measurement precision
If the MIP method uses maximum pixel value from three-dimensional data, then images with small contrast can be distinctly depicted, but information of pixel value between pixels is lost
Solution Approach 1:
The patent introduces frequency separation as an intermediary process between MIP image generation and final image output. This intermediary step extracts high-frequency components (edges and fine structures) and low-frequency components (overall intensity distribution) separately, allowing the system to preserve pixel value information that would otherwise be lost in the maximum intensity projection process.
Solution Approach 2:
The patent transitions from spatial domain processing to frequency domain processing by applying Fourier transform or wavelet transform. This dimensionality change allows simultaneous preservation of both contrast information (through high-frequency components) and overall intensity information (through low-frequency components), overcoming the information loss inherent in conventional MIP.
3Reliability
If a projection process is performed on three-dimensionally constructed data in any viewpoint direction, then synthetic mammogram images are generated with good noise resistance, but the image quality deteriorates due to using only maximum pixel value
Solution Approach 1:
The patent merges multiple processing results by combining the noise-resistant synthetic mammogram image from MIP processing with the detailed structural information extracted through frequency separation. The final image is constructed by integrating high-frequency components from the MIP-based synthetic image with low-frequency components, thereby achieving both noise resistance and high image quality.
Solution Approach 2:
The patent creates a composite image structure by combining different frequency components from the projection data. Rather than relying solely on maximum pixel values, the system constructs a composite representation that includes both high-frequency edge information and low-frequency intensity distribution, resulting in an image that兼具 noise resistance and diagnostic quality.
Data Source
AI summary
A control unit sets an angle of incidence of radiation, incident on a detection surface of a radiation detector with respect to the detection surface, to a plurality of angles different from each other including an angle (0 degrees) in a normal direction of the detection surface, and acquires at least one of a plurality of projection images captured by the radiation detector in accordance with the angle of incidence or reconstructed images reconstructed using the projection images. In addition, the control unit generates a first synthetic mammogram image on the basis of the one image. In addition, the control unit 40 generates a second synthetic mammogram image by synthesizing a high-frequency image having a high-frequency component higher than a predetermined frequency of the first synthetic mammogram image and a low-frequency image having a low-frequency component of equal to or lower than a predetermined frequency of a zero-degree projection image.