Nuclear Image Standardization via Voxel Distribution Templates

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Nuclear medicine images produced by different hospitals and instruments vary in quality, leading to inconsistent interpretations and limitations in using machine learning for automated image analysis due to subjective human interpretation and variability in image data.

Innovation Solution

A nuclear image processing method that standardizes images using a data augmentation algorithm to generate image standardization templates, enabling cross-hospital and cross-instrument consistency, and enhances image quality for improved machine learning and automated interpretation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual interpretation by nuclear medicine specialists is used, then interpretation can be performed, but subjective experience leads to inconsistent results and high possibility of misjudgment

Engineering Contradiction:
Improveinterpretation consistencyVSAvoidmanual interpretation dependency
Core Design Contradiction:
ReliabilityVSExtent of automation

Solution Approach 1:

The system enables self-service interpretation through automated image processing and analysis algorithms that independently process nuclear images without requiring manual specialist intervention for each case, thereby reducing subjective variability while maintaining interpretation capability

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical system of manual human interpretation with an automated computational system that uses standardized processing algorithms to analyze images, substituting human subjective judgment with objective machine-based analysis

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If images from different hospitals and instruments are used, then more data is available, but image quality varies leading to inconsistent interpretation results

Engineering Contradiction:
Improvecross-hospital compatibilityVSAvoidimage quality consistency
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The system changes key parameters including image resolution, contrast levels, and spatial normalization to establish uniform standards across images from different sources, transforming variable input images into standardized formats that ensure consistent interpretation results

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent creates a universal image processing framework that can handle images from multiple hospitals and instrument types through standardized protocols, making the system adaptable to diverse inputs while maintaining consistent output quality

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

3Manufacturing precision

If data augmentation algorithm is applied to generate standardization templates, then image standardization is improved, but processing complexity increases

Engineering Contradiction:
Improveimage standardization accuracyVSAvoidprocessing algorithm complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by pre-generating standardization templates through data augmentation algorithms during an offline training phase, so that during actual image processing, only simple template matching and adjustment are needed, reducing real-time processing complexity

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS11311261B2Nuclear image processing method
Publication Date: 2022.04.26 ATOMIC ENERGY COUNCIL INSTITUTE OF NUCLEAR ENERGY RESEARCH
  • US11311261B2 patent drawing
  • US11311261B2 patent drawing
  • US11311261B2 patent drawing

AI summary

A nuclear image processing method is provided. The method includes the following steps: inputting a normalized standard space nuclear image; selecting a voxel of the normalized standard space nuclear image and collecting the values of the neighbor voxels to form a voxel value set; conducting a data augmentation algorithm to generate a voxel distribution function; calculating an expected value of the distribution and calculating a first standard deviation of the portion over the expected value and a second standard deviation of the portion lower than the expected value; repeating the above steps to calculate the expected value, the first standard deviation and the second standard deviation of the necessary voxels, so as to form an image standardization template set including expected value template, first standard deviation template and the second standard deviation template.