Image Luminance Histogram Matching for Cross-CT Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis techniques using AI face challenges in achieving accurate machine learning due to differences in imaging conditions across various CT imaging apparatuses, leading to variations in image appearance and reduced model accuracy.
Innovation Solution
An image processing method that generates a conversion rule to align the luminance value histograms of images captured under different conditions, allowing for improved similarity and accuracy in machine learning by converting pixel luminance values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If images are collected from multiple CT imaging apparatuses with different versions and settings, then the quantity of training images increases, but the similarity among imaging conditions deteriorates leading to reduced machine learning accuracy
Solution Approach 1:
The patent applies parameter changes by adjusting luminance values of images through conversion rules. These rules are generated by comparing histograms of luminance values between images from different imaging apparatuses and modifying parameters such as luminance conversion coefficients to align the appearance characteristics of images across different devices, thereby maintaining machine learning accuracy while utilizing diverse training data
2Reliability
If common preprocessing is applied to reduce difference in image appearance, then image similarity improves, but the complexity of image processing increases
Solution Approach 1:
The patent implements preliminary action by pre-calculating conversion rules based on histogram comparisons between different imaging apparatuses. These conversion rules are prepared in advance and stored, allowing for efficient application during actual machine learning without requiring complex real-time processing, thus reducing operational complexity while maintaining image appearance consistency
Solution Approach 2:
The patent uses parameter changes by applying luminance conversion rules that adjust image parameters based on pre-computed conversion factors. This approach simplifies the preprocessing operation to straightforward parameter transformation rather than complex image processing algorithms, reducing computational complexity while achieving image appearance alignment
Data Source
AI summary
An information processing apparatus generates a histogram indicating the number of pixels for an individual luminance value of a first image obtained by capturing an image of a first subject under a first imaging condition, generates a histogram indicating the number of pixels for an individual luminance value of a second image obtained by capturing an image of a second subject of a same type as the first subject under a second imaging condition, generates a conversion rule for the luminance values of the pixels of the second image to improve a similarity between the histograms of the first image and the second image, converts a luminance value of an individual pixel of a third image obtained by capturing an image of a third subject of the same type under the second imaging condition, by using the conversion rule, and executes machine learning by using the third image.