Convolutional Neural Network for Low-Dose PET Image Correction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current PET/CT imaging systems require standard PET doses and combined CT or MRI scans for clinical-quality image reconstruction, leading to increased patient exposure, longer scan times, and lower throughput, while also being costly due to the need for full detector configurations.

Innovation Solution

A computer-implemented method using a convolutional neural network to reconstruct images from low-dose PET data, correcting for scatter and attenuation without the need for CT or MRI scans, by training the network with standard-dose PET sinogram data and its corrections, enabling the generation of fully corrected standard-dose equivalent images from low-dose data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If standard PET dose and combined CT or MRI scans are used for image reconstruction, then image quality is improved, but patient radiation exposure increases

Engineering Contradiction:

Improveimage qualityVSAvoidpatient radiation exposure

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

A neural network is introduced as an intermediary between low-dose PET data and corrected images. The network learns the complex relationship between low-dose and standard-dose images through training, acting as a mediator that performs attenuation and scatter correction without requiring actual CT or MRI scans. This intermediary model enables quality image reconstruction while avoiding the additional radiation from companion imaging.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates a learned copy of the correction process by training the neural network on pairs of low-dose and standard-dose images. Instead of directly using CT/MRI data for correction, the network learns to simulate the correction effect, producing images that are equivalent to fully corrected standard-dose images but from low-dose input alone.

Inventive Principle:

Principle #26Copying

2Measurement precision

If standard PET dose and combined CT or MRI scans are used for image reconstruction, then image quality is improved, but scan time increases

Engineering Contradiction:

Improveimage qualityVSAvoidscan time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and removes the CT or MRI scanning step from the conventional PET imaging workflow. By training a neural network to perform the correction functions that traditionally required separate anatomical scans, the method eliminates these time-consuming companion imaging procedures while maintaining image reconstruction quality.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The neural network is pre-trained offline on large datasets of paired low-dose and standard-dose images. This preliminary training phase captures the complex correction relationships, enabling rapid real-time correction during actual patient scanning without requiring the time-consuming CT or MRI acquisition steps.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If standard PET dose and combined CT or MRI scans are used for image reconstruction, then image quality is improved, but imaging throughput decreases

Engineering Contradiction:

Improveimage qualityVSAvoidimaging throughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent removes the bottleneck of companion CT or MRI scanning from the imaging workflow. By replacing these sequential scanning steps with a pre-trained neural network that processes PET data independently, the system eliminates the time loss associated with scheduling and performing separate anatomical scans, thereby increasing overall imaging throughput.

Inventive Principle:

Principle #2Taking out (Extraction)

4Measurement precision

If full detector configurations are used in PET scanners, then image quality is improved, but system cost increases

Engineering Contradiction:

Improveimage qualityVSAvoidsystem cost

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The neural network serves as an intermediary that compensates for the limitations of sparse detector configurations. By learning the patterns and missing information from training data, the network fills in gaps caused by fewer detectors, enabling acceptable image quality reconstruction without requiring expensive full-configuratio ndetector arrays.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20220130079A1Systems and methods for simultaneous attenuation correction, scatter correction, and de-noising of low-dose pet images with a neural network

Publication Date: 2022.04.28 SIEMENS MEDICAL SOLUTIONS USA INC

AI summary

An image reconstruction system generates de-noised, attenuation corrected, and scatter corrected images using AI processing. The system receives a low-dose PET image and applies a machine learning algorithm via a convolutional neural network to the low-dose PET image to generate an output image. The output image includes correction for scatter and attenuation associated with the image being low-dose. The system provides the output image to a computing device comprising a user interface.