Deep Learning Reconstruction for Limited-Angle CBCT Perfusion Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing perfusion imaging methods face challenges in improving sampling count, temporal resolution, and image quality in cone beam computed tomography (CBCT) perfusion imaging, particularly in the reconstruction of limited-angle images, which are prone to artifacts.
Innovation Solution
A medical imaging method utilizing a deep learning model with a series of primary and secondary feature extraction blocks to process limited-angle reconstructed images, enhancing image quality by iteratively optimizing features and removing artifacts through a reverse-order processing approach.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional CBCT reconstruction methods are used for limited-angle imaging, then the reconstruction process is simple and fast, but image quality deteriorates due to artifacts
Solution Approach 1:
A deep learning model is introduced as an intermediary between the limited-angle projection data and the final reconstructed image. The model processes the initial reconstruction and removes artifacts, thereby improving image quality without requiring changes to the basic CBCT hardware or reconstruction pipeline
Solution Approach 2:
The patent replaces traditional iterative reconstruction algorithms (which are computationally intensive and complex) with a trained deep learning model. Once trained, the model provides fast artifact removal with simpler computational requirements during inference, trading off training complexity for operational simplicity
2Manufacturing precision
If more projection data is collected to improve image quality, then image quality improves, but scanning time increases
Solution Approach 1:
The patent applies partial action by collecting limited-angle projection data (less than a full 360-degree scan) and using a deep learning model to compensate for the missing data. This approach achieves acceptable image quality with reduced scanning time by not requiring complete projection data coverage
Solution Approach 2:
The patent extracts and removes artifacts from the reconstructed image using a deep learning model. By specifically targeting and eliminating artifact components rather than requiring complete raw data, the system achieves improved image quality without the time penalty of extended scanning
3Manufacturing precision
If artifact removal techniques are applied to improve image quality, then image quality improves, but processing time increases
Solution Approach 1:
The deep learning model is trained in advance on a large dataset of paired images (with and without artifacts). This preliminary training phase allows the model to learn artifact patterns and removal strategies, so that during actual operation, artifact removal occurs rapidly during the inference phase without significant processing delays
Data Source
AI summary
The present disclosure provides a system and method for medical imaging. The method includes obtaining a first reconstructed image of a target subject; and generating a target reconstructed image of the target subject by inputting the first reconstructed image into a trained deep learning model. An image quality of the target reconstructed image is higher than that of the first reconstructed image.