Deep Learning Generation of Photon-Counting Spectral Data from Dual-Energy CT

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The high cost and complexity of photon-counting CT systems, along with the need for specialized imaging protocols, have hindered their adoption in clinical settings, despite their potential for improved material quantification and reduced noise in imaging.

Innovation Solution

A deep learning regression algorithm is employed to generate photon-counting spectral image data from non-photon counting X-ray spectral data using dual-energy CT data, leveraging inter-voxel local statistics to predict photon-counting results without requiring specialized hardware or new protocols.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If photon-counting CT systems are used, then material quantification and noise reduction are improved, but cost and device complexity increase

Engineering Contradiction:
Improvematerial quantificationVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent uses a deep learning regression algorithm to generate synthetic photon-counting spectral image data from non-photon-counting spectral data. Instead of requiring actual photon-counting hardware, the system creates a virtual copy of photon-counting results through computational modeling, thereby achieving photon-counting quality images without the associated hardware complexity and cost.

Inventive Principle:
Principle #26Copying

2Reliability

If photon-counting CT systems are used, then noise in imaging is reduced, but cost increases

Engineering Contradiction:
Improvenoise reductionVSAvoidsystem cost
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system generates synthetic photon-counting spectral data that replicates the noise-reduction characteristics of actual photon-counting systems. By computing virtual photon-counting results from conventional spectral data through deep learning, the method achieves the reliability benefits without requiring expensive photon-counting detector hardware.

Inventive Principle:
Principle #26Copying

3Measurement precision

If specialized imaging protocols are required, then photon-counting results are achieved, but ease of operation decreases

Engineering Contradiction:
Improvephoton-counting resultsVSAvoidimaging protocol complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The deep learning regression algorithm is trained to handle multiple types of non-photon-counting spectral data inputs and generate photon-counting results across different imaging scenarios. This universal approach allows the system to achieve photon-counting quality results using existing conventional imaging protocols, eliminating the need for specialized acquisition procedures while maintaining measurement precision.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4076198B1Apparatus for generating photon counting spectral image data
Publication Date: 2025.10.15 KONINKLIJKE PHILIPS NV
  • EP4076198B1 patent drawingFigure 1~3
  • EP4076198B1 patent drawingFigure 4
  • EP4076198B1 patent drawingFigure 5

AI summary

The present invention relates to an apparatus (10) for generating photon counting spectral image data, comprising: an input unit (20); a processing unit (30); and an output unit (40). The input unit is configured to receive non-photon counting X-ray spectral energy data. The processing unit is configured to implement a deep learning regression algorithm to generate photon counting X-ray spectral data, and the generation comprises utilization of the non-photon counting X-ray spectral energy data. The output unit is configured to output the photon counting X-ray spectral data.