CT Detector Data Compression for Low-Loss Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing CT data compression methods are inflexible and require specific adaptation to subsequent applications, leading to information loss and sub-optimum performance across different usage cases, particularly in high-resolution spectral reconstructions.
Innovation Solution
Employing a trained machine learning model (MLM) to compress and decompress CT data using a compression module and decompression module, respectively, tailored to the specific characteristics of X-ray detector data, including intensity, count events, and spatial correlations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional compression methods (transfer functions, hybrid data coding, temporal compression) are used, then data bandwidth is reduced, but the compression is sub-optimum for specific applications and information loss occurs
Solution Approach 1:
The patent applies parameter changes by transforming the data representation from raw detector signals to compressed representations using learned parameters. The machine learning model learns optimal parameter transformations that preserve essential information while reducing data dimensionality, enabling adaptive compression that maintains reconstruction quality across different CT applications.
Solution Approach 2:
The patent replaces traditional mechanical compression systems (transfer functions, hybrid data coding) with a machine learning-based system. The neural network automatically learns compression and decompression transformations, substituting manual algorithm design with data-driven learning that adapts to the specific characteristics of CT data and application requirements.
2Manufacturing precision
If compression is adapted to specific applications, then reconstruction quality improves, but the system loses flexibility for other applications
Solution Approach 1:
The patent implements universality by training a single machine learning model to handle multiple CT applications with different resolution and spectral requirements. The model learns a general compression representation that can be decompressed to serve various purposes (standard CT, high-resolution reconstruction, spectral analysis) without requiring separate compression systems for each application type.
Solution Approach 2:
The patent applies dynamics by enabling the compression system to adapt its behavior based on application requirements. The machine learning model can dynamically adjust the compression and decompression transformations according to the specific needs of different CT applications, providing flexible quality adjustment without requiring manual reconfiguration of compression parameters.
3Adaptability or versatility
If a trained machine learning model is used for compression, then adaptability to different applications improves, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training the machine learning model during an offline phase before actual CT data processing. The model learns optimal compression and decompression transformations during training using simulated or historical data, so that during actual operation the model can quickly process new data without requiring complex real-time learning computations.
Solution Approach 2:
The patent uses copying by creating a trained model that encapsulates learned compression knowledge. Instead of performing complex learning computations during data processing, the system copies the learned transformations from the trained model to compress new CT data, significantly reducing real-time computational requirements while maintaining adaptability.
Data Source
AI summary
In order to compress data from a system for CT, which has an X-ray detector containing a detector pixel array, a first training dataset is obtained, which contains a pixel value for each of a first multiplicity of detector pixels of the detector pixel array, which pixel value relates to an intensity of X-ray radiation incident on the detector pixel concerned. The first dataset is compressed by applying a first compression module to first input data, which contains the first dataset. The first compression module is comprised by a trained first machine learning model, MLM, which is trained to compress input data via the first compression module, and to reconstruct at least some of the input data based on the compressed input data via a first decompression module of the first MLM.