Medical Image Super-Resolution by Sensor Mode Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing superresolution processing methods using convolutional neural networks (CNN) for medical images fail to maintain image quality when the generation method of low-resolution images differs from the method used during training, leading to reduced resolution and loss of detailed information.
Innovation Solution
A medical-image processing apparatus that selects a learning model based on the operation mode of the sensor used for image capture, generating a medical image of higher resolution by matching the input image's generation method with the method used in training, utilizing a CNN with parameters adjusted for the specific sensor operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If multiple pixels are operated as one pixel to reduce radiation exposure, then radiation exposure to the examinee is reduced, but resolution and detailed information in the radiographic images deteriorate
Solution Approach 1:
The system performs preliminary classification of the low-resolution image generation method before applying super-resolution processing. By identifying whether the input image was generated through pixel grouping, binning, or other methods, the system selects the appropriate learning model trained for that specific method, thereby optimizing resolution recovery while maintaining the radiation reduction benefit
Solution Approach 2:
The system changes the parameter of the learning model selection based on the generation method parameter of the input low-resolution image. Different learning models are configured with parameters optimized for different sensor operation modes, allowing the system to adaptively adjust the super-resolution processing parameters to match the input image characteristics
2Productivity
If a CNN trained on one low-resolution image generation method is used to process images generated by another method, then processing speed is maintained, but super-resolution image quality deteriorates
Solution Approach 1:
The system segments the super-resolution processing into multiple specialized learning models, each trained on a specific low-resolution image generation method. This segmentation allows each model to be optimized for its specific task while maintaining overall processing efficiency through automated model selection based on input image characteristics
Solution Approach 2:
The system creates a universal super-resolution processing framework that handles multiple low-resolution image generation methods through a single integrated system. The framework includes multiple learning models that can be selectively applied based on the input image type, providing multi-functionality while maintaining specialized performance for each method
Data Source
AI summary
A medical-image processing apparatus according to the present invention includes an obtaining unit configured to obtain a medical image obtained by capturing an image of an examinee and a generation unit configured to input the medical image to a learning model selected based on an operation mode of a sensor at the image capturing to generate a medical image of a higher resolution than a resolution of the medical image.