CT Reconstruction Neural Network for High Pitch Artefact Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High relative pitch in CT scans leads to incomplete angular coverage, resulting in artefacts in image reconstruction, particularly affecting outer regions of the volume in the x/y plane.
Innovation Solution
An apparatus comprising an input unit, a processing unit, and an output unit, where the processing unit uses a machine learning algorithm, specifically a neural network, trained on CT slice reconstruction data with and without high relative pitch artefacts, to correct CT X-ray data acquired at high relative pitch and produce artefact-free or reduced artefact CT slice reconstruction data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If high relative pitch is used to scan large volume quickly, then productivity is improved, but image quality deteriorates due to reconstruction artefacts
Solution Approach 1:
A deep learning neural network is introduced as an intermediary between the raw projection data and the final reconstructed images. The network learns to map incomplete high-pitch projection data to complete low-pitch projection data, effectively mediating the information gap caused by high pitch scanning and enabling artifact-free reconstruction
Solution Approach 2:
The method creates a virtual copy of the incomplete high-pitch projection data by training the neural network to generate synthetic projection data that matches what would have been obtained at low pitch. This copied data is then used for reconstruction, allowing the system to achieve low-pitch image quality from high-pitch acquisitions
2Productivity
If dual source systems are used to overcome pitch limits, then productivity is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces the mechanical solution of using dual X-ray sources with a computational solution based on deep learning. Instead of adding physical hardware complexity to achieve high pitch scanning, the method uses a neural network to process and reconstruct data from a single source, achieving the same productivity improvement without the associated hardware complexity and cost
3Productivity
If detectors with larger z-coverage are used to enable high pitch, then productivity is improved, but device complexity and cost increase
Solution Approach 1:
The method changes the computational parameters of data processing rather than the physical parameters of the detector. By applying deep learning algorithms to process the existing detector data, the system achieves high-pitch scanning capability without modifying the detector's physical z-coverage, thereby avoiding increased device complexity and cost
Data Source
AI summary
The present invention relates to an apparatus (10) for correcting computer tomography (“CT”) X-ray data acquired at high relative pitch, the apparatus comprising: an input unit (20); a processing unit (30); and an output unit (40). The input unit is configured to provide the processing unit with CT X-ray data of a body part of a person acquired at high relative pitch. The processing unit is configured to determine CT slice reconstruction data of the body part of the person with no or reduced high relative pitch operation reconstruction artefacts using a machine learning algorithm. The machine learning algorithm was trained on the basis of CT slice reconstruction data, and wherein the CT slice reconstruction data comprised first CT slice reconstruction data with high relative pitch reconstruction artefacts and comprised second CT slice reconstruction data with less, less severe, or no high relative pitch reconstruction artefacts. The output unit is configured to output the CT slice reconstruction data of the body part of the person.