Mammographic Image Multi-Channel Conversion for Malignancy Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mammographic image processing technologies have suboptimal sensitivity and specificity in detecting malignancy, with traditional CAD algorithms failing to improve cancer detection accuracy despite widespread implementation.
Innovation Solution
A method and system that convert single channel 2D mammographic images into multi channel images using Contrast Limited Adaptive Histogram Equalization (CLAHE) and employ a trained statistical classifier with sub-classifiers for whole image and patch analysis, along with a gating component to compute likelihood and location of malignancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional CAD algorithms are implemented in mammographic image processing, then automation extent is improved, but measurement precision of malignancy detection deteriorates
Solution Approach 1:
The mammographic image is divided into multiple patches that are processed independently by separate neural networks. Each patch is analyzed for malignancy indicators, and the results are aggregated to form a comprehensive diagnosis. This segmentation allows the system to handle complex detection tasks with specialized sub-networks, improving overall detection precision while maintaining automation.
Solution Approach 2:
The system transforms the detection problem from analyzing entire images globally to analyzing local patches with focused attention. By examining multiple localized regions independently and then integrating results, the system achieves higher measurement precision in identifying malignancy indicators while maintaining full automation through the neural network ensemble.
2Measurement precision
If multiple neural networks process entire mammographic images, then measurement precision is improved, but loss of time in processing increases
Solution Approach 1:
By dividing the large mammographic image into smaller patches, the system enables parallel processing of multiple regions simultaneously. Each neural network processes a specific patch independently, reducing the computational burden on individual networks and allowing faster aggregation of results compared to processing entire images sequentially through multiple networks.
Solution Approach 2:
The system processes only relevant local patches rather than analyzing every pixel of the entire image in detail. By focusing computational resources on specific regions of interest identified through patch-based analysis, the system achieves high detection precision while minimizing unnecessary processing time on non-critical areas.
3Device complexity
If single channel 2D mammographic images are used, then device complexity is reduced, but measurement precision of malignancy detection deteriorates
Solution Approach 1:
The system transforms single-channel 2D images into multi-channel representations by creating multiple processed versions of the input image with different enhancements. Each channel highlights different features or aspects of the mammographic data, allowing neural networks to detect malignancy indicators more effectively while maintaining relatively simple input requirements.
Solution Approach 2:
The system applies various image processing transformations to create multiple channels from the single input image, changing parameters such as contrast, brightness, and frequency domain representations. These parameter transformations enhance the detectability of malignancy indicators without requiring complex multi-channel input devices, thus maintaining low device complexity while improving measurement precision.
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
Improves detection accuracy and efficiency by enhancing sensitivity and specificity to levels comparable to expert radiologists, reducing processing time and data storage requirements, and effectively differentiating between benign and malignant breast tissue.
Implementation Method 1
converting the single channel 2D mammographic image into a multi channel 2D mammographic image using Contrast Limited Adaptive Histogram Equalization (CLAHE)
Data Source
AI summary
There is provided a method of computing a likelihood of malignancy in a mammographic image, comprising: receiving a single channel 2D mammographic image including a single pixel intensity value for each pixel thereof, converting the single channel 2D mammographic image into a multi channel 2D mammographic image including multiple pixel intensity value channels for each pixel thereof, computing by a first sub-classifier according to the whole multi channel image, a first score indicative of likelihood of malignancy within the whole multi channel image, computing by a second sub-classifier according to each respective patch extracted from the multi channel image, a respective second score indicative of likelihood of malignancy within each respective patch, and computing by a gating sub-classifier according to the first score and the second scores, an indication of likelihood of malignancy and a location of the malignancy.