Radiological Image Enhancement Using Fused Anomaly Influence Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radiological image processing methods require human intervention for parameterization and ROI definition, leading to increased diagnosis time and suboptimal anomaly visualization, and machine learning-based systems are difficult to interpret.
Innovation Solution
A method using a convolutional neural network to detect radiological anomalies, normalize and fuse influence maps, and weight pixel contributions in histogram processing to enhance anomaly visibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If post-processing operations are applied to the entire image, then the processing can be automated without user intervention, but the results are not optimum for visualizing anomalies in specific regions
Solution Approach 1:
The patent segments the image processing by dividing it into multiple regions of interest (ROIs) based on detected anomalies. Each ROI is processed independently with optimized parameters, rather than applying uniform processing to the entire image. This allows automated processing while achieving optimal visualization quality for each specific anomaly region.
Solution Approach 2:
The patent applies local quality enhancement by using adaptive histogram equalization and contrast enhancement techniques specifically tailored to each ROI. The processing parameters are locally optimized for each anomaly region, ensuring that each area receives the appropriate level of enhancement rather than a one-size-fits-all approach.
2Manufacturing precision
If a radiologist manually parameterizes post-processing operations or defines ROIs, then the processing can be optimized for specific anomalies, but the diagnosis time is increased
Solution Approach 1:
The patent performs preliminary anomaly detection using a convolutional neural network before the post-processing stage. This preliminary action automatically identifies the locations and characteristics of anomalies, which then guides the subsequent optimized processing. This eliminates the need for manual ROI definition while ensuring that processing is focused on the correct regions.
Solution Approach 2:
The system performs self-service by automatically detecting anomalies and generating optimized processing parameters without requiring radiologist intervention. The convolutional neural network autonomously identifies anomalies and the system automatically configures the post-processing parameters, making the entire workflow self-contained and eliminating manual time investment.
3Extent of automation
If machine learning algorithms are used to assist diagnosis, then the processing can be automated, but the results are difficult to interpret
Solution Approach 1:
The patent uses influence maps as an intermediary between the convolutional neural network's internal processing and the final output. These influence maps visually represent which regions of the input image most influenced the anomaly detection, providing interpretable information that bridges the gap between automated machine learning and radiologist understanding. The influence maps serve as a mediator that makes the black-box algorithm transparent.
Data Source
AI summary
A method for processing a radiological image, in digital format includes at least one radiological anomaly detected using a convolutional neural network that has been trained to detect radiological anomalies on radiological examinations, the radiological image being characterized by the intensity of each of its pixels, and by at least one radiological anomaly influence map that assigns, for each pixel of the radiological image, a value representative of the proportion to which the pixel had an influence on the detection result of the radiological anomaly, which method is computer-implemented and comprises the steps of: normalizing the radiological anomaly influence maps to give normalized radiological anomaly influence maps; fusing the normalized radiological anomaly influence maps to give a single fused influence map; carrying out improvement processing on the image, using an intensity histogram, wherein the contribution of each pixel in the computing of the intensity histogram is weighted by the fused influence map.