Acquisition-Aware Deep Neural Networks for Medical Image Deblurring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods struggle to effectively deblur medical images due to unknown blurring sources and varying extents of blurring, leading to degraded image resolution and diagnostic quality, and often result in sharper but noisier images.
Innovation Solution
A deep neural network system that incorporates acquisition parameters into the image processing, using a trained deep neural network and acquisition parameter transforms to map blurred medical images to sharper, consistent images by leveraging information about the type, extent, and distribution of blurring artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional deblurring methods are used on medical images, then image sharpness may be improved, but image noise increases significantly
Solution Approach 1:
The patent applies parameter changes by incorporating acquisition parameters (such as echo time, echo train length, and other imaging settings) into the deep neural network processing. These parameters are transformed and integrated as additional input channels to the network, allowing the model to adapt its deblurring operation based on the specific acquisition conditions. This enables the network to adjust its denoising strength and sharpness enhancement according to the actual blurring characteristics, thereby improving image quality without excessive noise amplification.
2Manufacturing precision
If deep learning techniques are applied to deblur medical images, then diagnostic quality can be improved, but computational complexity and processing time increase
Solution Approach 1:
The patent implements preliminary action by pre-processing the acquisition parameters through a dedicated acquisition parameter transform module before feeding them into the deep neural network. This transform converts the raw acquisition parameters into a standardized format that the network can efficiently process. Additionally, the network architecture is designed to integrate these transformed parameters early in the processing pipeline, allowing the model to leverage this information throughout subsequent processing stages without requiring excessive computational resources during inference.
3Measurement precision
If acquisition parameters are incorporated into the deep neural network, then blurring artifacts can be more accurately addressed, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary component called the acquisition parameter transform, which serves as a bridge between the raw acquisition parameters and the deep neural network. This transform module converts the acquisition parameters into a standardized representation that can be seamlessly integrated with the image data. By using this intermediary, the system avoids directly complicating the core neural network architecture while still enabling the network to utilize acquisition parameter information for more accurate blurring correction.
Data Source
AI summary
Methods and systems are provided for processing medical images using deep neural networks. In one embodiment, a medical image processing method comprises receiving a first medical image having a first characteristic and one or more acquisition parameters corresponding to acquisition of the first medical image, incorporating the one or more acquisition parameters into a trained deep neural network, and mapping, by the trained deep neural network, the first medical image to a second medical image having a second characteristic. The deep neural network may thereby receive at least partial information regarding the type, extent, and/or spatial distribution of the first characteristic in a first medical image, enabling the trained deep neural network to selectively convert the received first medical image.