Medical Image Noise Reduction via Deep Learning Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical imaging technologies often face high noise levels in images reconstructed from low-dose scans or scans using ultra-long/ultra-short half-life drugs, leading to suboptimal image quality and signal-to-noise ratios.
Innovation Solution
A method involving a computing device that processes imaging data to generate a first image, then creates two intermediate images – one with feature information and another with reduced noise – using machine learning models, which are combined to produce a target image with improved quality and signal-to-noise ratio.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If low-dose imaging is performed to reduce radiation exposure, then radiation safety is improved, but image noise increases and image quality deteriorates
Solution Approach 1:
The patent introduces an intermediary deep learning model that acts as a mediator between the low-dose input image and the final high-quality output image. This model learns the mapping relationship from low-dose to high-quality images through training on paired datasets, effectively decoupling the radiation dose from the image quality requirement. The intermediary model enables quality enhancement without requiring additional radiation exposure.
Solution Approach 2:
The patent performs preliminary actions by pre-training the deep learning model on large datasets of paired low-dose and high-quality images before actual imaging. This preliminary training phase allows the model to learn optimal noise reduction and quality enhancement strategies in advance, so that during actual low-dose imaging, the model can immediately apply learned transformations without requiring additional radiation or time-consuming processing during the imaging procedure itself.
2Manufacturing precision
If image processing operations are performed to reduce noise, then image quality is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent replaces traditional mechanical/mathematical image processing systems (such as iterative reconstruction algorithms, filtering operations, and optimization methods) with a deep learning-based neural network system. This substitution allows the complex noise reduction and quality enhancement tasks to be performed through learned patterns rather than explicit computational algorithms, significantly reducing processing complexity and time while maintaining or improving image quality.
3Productivity
If traditional image reconstruction methods are used, then processing speed is maintained, but noise reduction capability is insufficient
Solution Approach 1:
The patent replaces traditional mechanical/mathematical image processing systems (such as iterative reconstruction algorithms, filtering operations, and optimization methods) with a deep learning-based neural network system. This substitution allows the complex noise reduction and quality enhancement tasks to be performed through learned patterns rather than explicit computational algorithms, significantly reducing processing complexity and time while maintaining or improving image quality.
Solution Approach 2:
The patent fundamentally changes the processing approach from deterministic algorithmic operations to probabilistic learned transformations. By training the neural network on diverse datasets with varying noise levels and image characteristics, the model learns to adaptively adjust processing parameters based on the input image characteristics, achieving superior noise reduction while maintaining processing efficiency through single-pass inference rather than iterative computation.
Data Source
AI summary
The present disclosure relates to systems and methods for image processing. The methods may include obtaining imaging data of a subject, generating a first image based on the imaging data, and generating at least two intermediate images based on the first image. At least one of the at least two intermediate images may be generated based on a machine learning model. And the at least two intermediate images may include a first intermediate image and a second intermediate image. The first intermediate image may include feature information of the first image, and the second intermediate image may have lower noise than the first image. The methods may further include generating, based on the first intermediate image and at least one of the first image or the second intermediate image, a target image of the subject.