PET Image Reconstruction via Coordinate Transformation and Data Merging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional PET reconstruction algorithms face challenges with unevenly spaced line-of-response (LOR) data, particularly near the center of a PET ring, which can corrupt Poisson statistics and result in non-uniform signal noise ratio (SNR) in PET images.

Innovation Solution

A system and method for processing PET data that involves merging overlapping imaging data from multiple bed positions, transforming data between coordinate systems, and applying weight coefficients to composite images, ensuring uniform SNR and accurate reconstruction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional reconstruction algorithms process pre-processed PET data, then the data can be reconstructed into images, but the Poisson statistics characteristics are corrupted and SNR becomes non-uniform

Engineering Contradiction:
Improveimage reconstruction qualityVSAvoidPoisson statistics characteristics
Core Design Contradiction:
Manufacturing precisionVSReliability

Solution Approach 1:

The patent applies preliminary action by performing coordinate system transformation and merging operations on raw PET data before reconstruction. The method transforms unevenly spaced LOR data into evenly spaced data in a new coordinate system, and merges overlapping data from multiple bed positions with appropriate weighting, thereby preparing the data in advance to preserve Poisson statistics and ensure uniform SNR throughout the field of view.

Inventive Principle:
Principle #10Preliminary action

2Length of moving object

If PET scanning covers a large axial length, then more of the patient can be examined, but the SNR becomes non-uniform across different regions

Engineering Contradiction:
Improveaxial coverage lengthVSAvoidSNR uniformity
Core Design Contradiction:
Length of moving objectVSManufacturing precision

Solution Approach 1:

The patent applies merging by combining PET data from multiple bed positions with overlapping regions. Data from adjacent bed positions are merged with weight coefficients that account for the overlap, ensuring that regions covered by multiple bed positions contribute more reliably to the final image. This merging process maintains uniform SNR across the entire axial coverage while extending the scanable length.

Inventive Principle:
Principle #5Merging (Combining)

3Shape

If LORs are unevenly spaced near the center of the PET ring, then the detector geometry is maintained, but the data spacing becomes irregular affecting reconstruction accuracy

Engineering Contradiction:
Improvedetector ring geometryVSAvoiddata spacing uniformity
Core Design Contradiction:
ShapeVSMeasurement precision

Solution Approach 1:

The patent applies dimensionality change by transforming data from the polar coordinate system (natural for ring geometry) to a Cartesian coordinate system. This transformation maps unevenly spaced LOR data in the polar domain to evenly spaced data in the Cartesian domain, thereby preserving the detector's ring geometry while achieving uniform data spacing suitable for standard reconstruction algorithms.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3234910B1System and method for image reconstruction
Publication Date: 2025.05.14 SHANGHAI UNITED IMAGING HEALTHCARE
  • EP3234910B1 patent drawingFigure 1
  • EP3234910B1 patent drawingFigure 2~3
  • EP3234910B1 patent drawingFigure 4

AI summary

A system and method relating to image processing are provided. The method may include the following operations. First data at a first bed position and second data at a second bed position may be received. The first bed position and the second bed position may have an overlapping region. A first image and a second image may be reconstructed based on the first data and the second data, respectively. Third data and fourth data corresponding to the overlapping region may be extracted from the first data and the second data, respectively. Merged data may be generated by merging the third data and the fourth data. A third image may be reconstructed based on the merged data. A fourth image may be generated through image composition based on the first image, the second image, and the third image.