Medical Image Processing Apparatus for Lesion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for identifying lesions in medical images, such as cerebral infarction, require manual expertise and have low determination accuracy, with semi-automatic methods using thresholds or machine learning achieving only 60-80% accuracy and failing to accurately differentiate between bright and dark components in brain images.
Innovation Solution
An image processing apparatus that sets inner and peripheral regions in medical images, calculates intensity value frequency distributions, computes probability differences, and applies a detection range setting algorithm to highlight lesion sites accurately, incorporating machine learning for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If semi-automatic lesion determination using predetermined threshold or machine learning is used, then automation level increases, but determination accuracy remains low at 60-80%
Solution Approach 1:
The image is divided into multiple components based on intensity values, with each component processed separately to identify lesion sites. This segmentation allows the system to handle different tissue types and lesion characteristics independently, improving overall accuracy while maintaining automation.
Solution Approach 2:
Different processing parameters and detection ranges are applied to different intensity value components rather than using a single global threshold. This local quality approach enables accurate lesion detection in both bright and dark components by tailoring the analysis to each region's specific characteristics.
2Device complexity
If a single threshold is used for lesion determination, then processing simplicity is maintained, but accuracy in differentiating bright and dark components decreases
Solution Approach 1:
The system dynamically adjusts detection parameters based on the intensity value components identified in the image. Rather than using a fixed threshold, the detection range is adapted to each component's characteristics, enabling accurate differentiation of both bright and dark lesions while maintaining systematic processing.
Solution Approach 2:
The invention changes the detection parameter (intensity value range) according to the specific component being analyzed. By setting different detection ranges for different intensity value components, the system can accurately identify lesions across the full spectrum of image intensities without requiring overly complex multi-threshold processing.
Data Source
AI summary
An image processing apparatus includes: an inner region setting unit and a peripheral region setting unit setting, for an inner region in a medical image, a peripheral region in the medical image; an intensity value group probability distribution calculation unit calculating an inner region histogram for the inner region and calculating a peripheral region histogram for the peripheral region; a probability difference calculation unit calculating a probability difference distribution by calculating a difference value between the inner region histogram and the peripheral region histogram for each predetermined intensity value; a component image processing unit generating a component image based on the medical image; a detection range setting unit setting a detection range to each of the component images; a mask setting unit selecting a pixel to be highlighted for each of the component images based on the detection range; and a display processing unit outputting the highlighted component image.