Mammographic Image Enhancement via Local Histogram Revision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mammography techniques often miss breast cancer due to human errors such as poor perception or interpretation by radiologists, leading to missed diagnoses, especially when lesions are subtle or not clearly visible.
Innovation Solution
A computer-aided detection method that enhances mammographic images by generating a localized histogram for each pixel using an enhancement sliding window, revising pixel values, and storing them in a second image to improve the detection of breast abnormalities, thereby reducing human error in cancer detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mammography is performed to capture breast tissue images, then breast cancer detection is enabled, but human errors in perception and interpretation lead to missed diagnoses
Solution Approach 1:
The patent introduces an automated image analysis system as an intermediary between the mammography imaging process and radiologist interpretation. This system processes mammographic images through multiple algorithms including noise reduction, edge detection, and lesion characterization to provide objective analysis results that assist radiologists, thereby reducing human error in perception and interpretation while maintaining the essential role of radiologist expertise
Solution Approach 2:
The patent replaces the purely manual mechanical process of radiologist image interpretation with an automated computational system. The system uses computer algorithms to perform initial image processing, lesion detection, and characterization, substituting human sensory and cognitive processes with mechanical computation to eliminate fatigue, inattention, and experience-level variations that affect radiologist reliability
2Measurement precision
If image enhancement is performed using traditional methods, then processing speed is maintained, but subtle cancerous features are not sufficiently highlighted
Solution Approach 1:
The patent segments the image processing task into multiple distinct stages: noise reduction, contrast enhancement, edge detection, and lesion characterization. Each stage processes specific features independently using optimized algorithms, allowing parallel computation that maintains high processing speed while achieving superior lesion visibility through cumulative refinement at each segmentation step
Solution Approach 2:
The patent transitions from traditional 2D image enhancement to a multi-dimensional processing approach that considers spatial relationships, texture patterns, and intensity distributions across multiple scales. By analyzing images in additional dimensional spaces (such as frequency domain transformations and multi-scale wavelet decompositions), the system highlights subtle cancerous features that are invisible in standard 2D representations while maintaining computational efficiency through optimized mathematical operations
Data Source
AI summary
Detecting breast abnormalities includes receiving a first mammographic image having original pixels. A second mammographic image is generated by enhancing the first mammographic image. Enhancing the first mammographic image includes performing the following for each original pixel in at least a subset of the original pixels. A histogram is generated for a region surrounding the original pixel, the region defined by an enhancement sliding window. Using the histogram, a value of the original pixel is revised to obtain a revised value, and the revised value is stored in the second mammographic image. A breast abnormality location is detected based on the second mammographic image.