Learning Model Region Extraction Using Pseudo Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In medical imaging, especially in emerging countries, it is challenging to acquire clear medical images due to inadequate imaging environments and poor imaging techniques, making it difficult to extract regions of interest for diagnosis.
Innovation Solution
An information processing apparatus and method that utilize a learning model to perform learning using pseudo images generated by altering pixel values of original images, allowing for the extraction of regions of interest even from unsuitable images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If clear medical images are required for diagnosis, then diagnostic accuracy is improved, but image acquisition becomes difficult in emerging countries with inadequate imaging environments
Solution Approach 1:
The system performs preliminary image enhancement processing on unclear medical images before diagnosis. By pre-processing the images to improve their quality and make regions of interest more detectable, the system enables diagnosis to proceed even when the original images are of poor quality, thus resolving the contradiction between requiring clear images and the inability to acquire them in emerging countries
Solution Approach 2:
The system introduces an intermediary processing layer between image acquisition and diagnosis. This intermediary layer includes image enhancement algorithms that act as a bridge, transforming unclear images into enhanced versions that can be used for diagnosis, thereby enabling the system to adapt to poor imaging conditions without sacrificing diagnostic capability
2Adaptability or versatility
If region of interest extraction is performed on unclear images, then diagnostic capability is maintained, but extraction accuracy deteriorates
Solution Approach 1:
The system performs preliminary image enhancement as a prerequisite step before region of interest extraction. By enhancing the image quality in advance, the subsequent extraction process can achieve better accuracy than would be possible on the original unclear image, thus maintaining both adaptability to unclear images and extraction accuracy
Solution Approach 2:
The system maintains a continuous pipeline where image enhancement is followed by region extraction. This continuous processing ensures that the benefits of enhancement are fully utilized in the extraction step, preventing the useful action of enhancement from being wasted and ensuring accurate extraction even from originally unclear images
3Reliability
If learning models are trained on clear images only, then model performance is optimized, but utilization of unclear images is prevented
Solution Approach 1:
The system performs preliminary image enhancement on unclear images before they are input to the learning model. This preliminary action transforms the unclear images into enhanced versions that the model can process effectively, allowing the model to maintain its high performance while also being able to handle unclear images from emerging countries
Solution Approach 2:
The image enhancement process acts as an intermediary between the learning model and unclear images. The enhanced images serve as a bridge, allowing the model to receive input that is suitable for its processing while still representing the original unclear images, thus maintaining both model performance and adaptability
Data Source
AI summary
An information processing apparatus including at least one processor, wherein the processor is configured to cause a learning model that is used to extract a region of interest from an input image, to perform learning in response to an input of a pseudo image generated in a case where a pixel value of at least a part of an original image obtained by imaging a subject is changed, as data for learning.