Medical Image Quality Assessment Using Reference-Based Comparative Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Assessing medical image quality is challenging due to its subjective nature, making it difficult to determine if images are of sufficient quality for diagnosis, especially for technically difficult patients.
Innovation Solution
A method that uses both an absolute image quality value and a comparative image quality value, determined by processing medical images with machine-learning models, to assess the quality of medical images. The comparative value is obtained by comparing the image to a selected reference image, chosen based on analysis of the medical image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automatic image quality scoring is used with machine learning models, then image quality assessment becomes automated, but the subjectivity and challenge of annotating training data remains
Solution Approach 1:
The patent introduces a reference image as an intermediary element to bridge the gap between automated assessment and human judgment. The reference image serves as a mediator that enables comparative evaluation, allowing the system to assess image quality relative to a standard without requiring complex subjective annotation of absolute quality values. This intermediary approach simplifies the training data annotation process while maintaining automated assessment capabilities.
2Measurement precision
If absolute image quality values are used, then a single quality score is obtained, but the assessment is highly subjective and unreliable
Solution Approach 1:
Instead of attempting to assign absolute quality values directly (the conventional approach), the patent inverts the problem by using comparative assessment. Rather than asking 'how good is this image absolutely?', the system asks 'how does this image compare to a reference image?'. This inversion transforms the subjective absolute scoring problem into a more reliable comparative evaluation, where the reference image provides an objective baseline for comparison.
Solution Approach 2:
The patent changes the parameter being measured from absolute quality score to comparative quality difference. By shifting from measuring image quality on an absolute scale to measuring the difference between image and reference image, the system transforms a subjective measurement into a more objective and reliable assessment. This parameter transformation enables consistent evaluation across different images and practitioners.
3Reliability
If multiple images are acquired for technically difficult patients, then sufficient quality images may be obtained, but examination time and resources increase
Solution Approach 1:
The patent implements a feedback mechanism where the image quality assessment result is immediately available to guide the examination process. By providing rapid feedback on image quality using the reference-based comparative method, the system enables real-time decision-making about whether additional images are needed, preventing unnecessary repeated acquisitions and optimizing examination time while ensuring sufficient image quality.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
Embodiments of the invention aim to provide a method for assessing the quality of a medical image. This is achieved by determining, using machine learning models, an absolute image quality value and a comparative image quality value of the medical image. The comparative image quality value describes the relative quality of the image in relation to a suitably selected reference image. An overall image quality value is then determined based on the absolute image quality value and the comparative image quality value.