In-Reconstruction Filtering for PET List Mode Iterative Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Iterative-based reconstruction algorithms in PET and SPECT imaging face challenges with noise accumulation leading to degraded image quality and reduced signal-to-noise ratio, particularly in low-count acquisitions, which can result in compromised diagnostic usefulness due to blurring and loss of quantitative accuracy when using post-reconstruction filters.

Innovation Solution

The implementation of an in-reconstruction filtering method, such as a median filter, during iterative updates in the image domain to control noise and maintain image resolution, allowing for higher iteration counts without degrading image quality, and improving quantitative accuracy in small objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If post-reconstruction filters are applied to smooth reconstructed images, then noise levels are reduced, but image resolution and quantitative accuracy are degraded

Engineering Contradiction:
Improvenoise levelsVSAvoidimage resolution and quantitative accuracy
Core Design Contradiction:
Object-affected harmful factorsVSManufacturing precision

Solution Approach 1:

The patent applies filtering operations during the iterative reconstruction process itself, rather than as a post-processing step. By incorporating filters into the forward and back projection operations during iteration, noise control is achieved before the final image is formed, preserving quantitative accuracy while reducing noise levels

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces filter operators as intermediary elements within the iterative reconstruction algorithm. These filter operators modify the forward projection and back projection operations, acting as mediators that control noise accumulation during the reconstruction process without directly filtering the final image

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If iterative-based reconstruction algorithms are used to improve image accuracy, then noise accumulates with each iteration, but image quality and signal-to-noise ratio are reduced

Engineering Contradiction:
Improveimage accuracyVSAvoidnoise accumulation
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent implements feedback mechanisms within the iterative reconstruction algorithm where filter operations are applied based on the current state of the reconstruction. The filtering strength and application are adjusted according to iteration number and image characteristics, creating a feedback loop that controls noise accumulation while maintaining convergence to an accurate solution

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent employs dynamic filtering where the filter parameters and application strategy change during the iterative process. Filtering is applied more aggressively in early iterations when noise accumulates rapidly, and reduced or modified in later iterations as the image converges, creating a dynamic balance between noise control and accuracy preservation

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3108450B1In-reconstruction filtering for positron emission tomography (PET) list mode iterative reconstruction
Publication Date: 2019.04.10 KONINKLIJKE PHILIPS NV
  • EP3108450B1 patent drawingFigure 1
  • EP3108450B1 patent drawingFigure 2
  • EP3108450B1 patent drawingFigure 3~4

AI summary

A system (10) and a method (100) iteratively reconstruct an image of a target volume of a subject. In each iteration of a plurality of iterations, an estimate image of the target volume (54) is forward projected (58) and compared (62) to received event data (44) to determine a discrepancy (64). The discrepancy (64) is back projected (66) and the back projection (68) updates (70) the estimate image (54). In at least one iteration, the estimate image (54) is filtered (52) in the image domain prior to being back projected.