Medical Image Reconstruction Parameters with Bi-Level Quality Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical imaging technologies rely heavily on operator-dependent adjustment of imaging parameters, leading to inconsistent image quality, with novice operators struggling to achieve high-quality results, and no-reference image quality metrics lacking correlation with perceived ultrasound B-mode image quality.
Innovation Solution
A bi-level approach using both no-reference and full-reference image quality metrics to iteratively modify imaging parameters, where no-reference metrics guide optimization in clinical contexts and full-reference metrics provide physical accuracy, ensuring high-quality medical imaging data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If operator-dependent adjustment of imaging parameters is used, then image quality can be improved by experienced operators, but the process becomes subjective and inconsistent across different operators
Solution Approach 1:
The system performs self-adjustment of imaging parameters by automatically analyzing the captured image and modifying parameters based on objective quality metrics, eliminating the need for manual operator intervention and ensuring consistent results across different users
Solution Approach 2:
The system implements a feedback loop where image quality metrics are calculated from the captured images, and these metrics are used to automatically adjust imaging parameters in subsequent captures, creating a closed-loop control system that continuously optimizes image quality
2Extent of automation
If no-reference image quality metrics are used, then automated workflow is enabled, but the metrics do not correlate well with perceived ultrasound B-mode image quality
Solution Approach 1:
The system introduces a trained machine learning model as an intermediary that bridges the gap between no-reference metrics and perceived image quality, translating automated measurements into predictions that correlate with human assessment while maintaining workflow automation
Solution Approach 2:
The machine learning model is pre-trained on a dataset of images with known quality assessments before deployment, allowing it to predict perceived image quality accurately from no-reference metrics without requiring manual calibration during actual use
3Measurement precision
If full reference image quality metrics are used, then high correlation with subjective judgement is achieved, but reference images are typically absent in clinical environments
Solution Approach 1:
The system creates a virtual reference model through machine learning that replicates the quality assessment capabilities of full-reference metrics without requiring actual reference images, enabling the use of FR-IQM accuracy in clinical settings where only no-reference conditions exist
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A mechanism for generating a set of imaging parameters used to reconstruct medical imaging data from raw or partially-processed medical imaging data. In an inner loop, the imaging parameter(s) are modified using one or more first image metrics produced using a no-reference image quality metric technique on medical imaging data produced using the imaging parameter(s). In an outer loop, the imaging parameter(s) are modified using one or more second image metrics produced using a full-reference image quality metric technique on comparison medical imaging data. The comparison medical imaging data is produced by processing raw or partially-processed reference medical imaging data using the imaging parameter(s).