Helical CT Reconstruction for Ultra-Fast-Pitch Artifact Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Single-source CT scanners face limitations in scanning speed and image quality when using ultra-fast-pitch scans due to insufficient data acquisition, leading to artifacts in reconstructed images, especially in exams involving significant patient motion.
Innovation Solution
A convolutional neural network (UFP-net) is trained using a customized loss function to mitigate image artifacts by incorporating local and non-local operators, suppressing location- and structure-dependent artifacts in ultra-fast-pitch CT data, enabling accurate image reconstruction on single-source scanners.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If helical pitch is increased to improve scanning speed, then productivity is improved, but image quality deteriorates due to insufficient data for accurate reconstruction
Solution Approach 1:
The patent changes the pitch parameter from conventional values (≤1.5) to ultra-fast-pitch values (>1.5) while using machine learning reconstruction to maintain image quality. The trained neural network adapts to the specific data characteristics of ultra-fast-pitch scans, enabling accurate reconstruction despite the increased pitch.
Solution Approach 2:
The patent replaces traditional iterative reconstruction algorithms with a trained machine learning model. The neural network learns optimal reconstruction patterns from training data and applies them to ultra-fast-pitch scans, substituting the mechanical iterative computation process with a learned mapping function that achieves faster, artifact-free reconstruction.
2Device complexity
If conventional reconstruction methods are used on ultra-fast-pitch data, then device complexity is minimized, but image quality deteriorates due to artifacts
Solution Approach 1:
The patent performs preliminary training of the machine learning model using regular-pitch scan data before deployment. This pre-training phase prepares the network to handle the specific artifacts and data patterns of ultra-fast-pitch scans, enabling it to correct artifacts during actual reconstruction without requiring complex real-time adjustments.
Solution Approach 2:
The trained machine learning model acts as an intermediary between the raw ultra-fast-pitch projection data and the final reconstructed images. It processes the artifact-prone data through learned transformations, mediating the reconstruction process to produce artifact-free images while maintaining relative simplicity in the overall system architecture.
Data Source
AI summary
Images are reconstructed from data acquired using an ultra-fast-pitch acquisition with a CT system. As an example, an ultra-fast-pitch acquisition mode in single-source helical CT (≥1.5) can be used to acquire data. A trained machine learning algorithm, such as a neural network, is used to reconstruct images in which artifacts associated with insufficient data acquired in the ultra-fast-pitch mode are reduced. An example neural network can include customized functional modules using both local and non-local operators, as well as the z-coordinate of each image, to effectively suppress the location- and structure-dependent artifacts induced by the ultra-fast-pitch mode. The machine learning algorithm can be trained using a customized loss function that involves image-gradient-correlation loss and feature reconstruction loss.