Radiology Quality Auditor Workload Reduction via Confidence Triage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated quality assessment algorithms for radiology images have limited accuracy, making them unsuitable for exclusive use in regulatory auditing tasks, which requires stringent quality assurance assessments due to increased regulatory oversight.
Innovation Solution
An apparatus and method that combines automated quality assessment with manual auditing by generating quality rating confidence values to select images for manual review, thereby enhancing the accuracy of the auditing process while reducing the workload on auditors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated quality assessment algorithms are used exclusively for regulatory auditing, then productivity increases, but measurement precision deteriorates due to limited accuracy of automated algorithms
Solution Approach 1:
The auditing process is segmented into two distinct stages: an automated screening stage that processes all images efficiently, and a manual verification stage that focuses only on images with low confidence scores. This segmentation allows the system to leverage the high throughput of automated algorithms while ensuring accurate assessment through human review of uncertain cases.
Solution Approach 2:
The confidence value generated by the automated algorithm serves as an intermediary metric that determines which images require manual review. This intermediary mechanism enables the system to intelligently triage images, directing only those with ambiguous quality assessments to human auditors, thereby optimizing the balance between automation and manual verification.
2Reliability
If more images are audited with shorter intervals to maintain regulatory compliance, then reliability of quality assurance improves, but workload on auditors increases
Solution Approach 1:
The automated quality assessment algorithm performs self-service by pre-evaluating all images and identifying those that require manual review. This self-service capability reduces the burden on auditors by automatically handling the initial screening, allowing them to focus only on images with low confidence scores rather than reviewing every image manually.
Solution Approach 2:
The automated algorithm performs preliminary action by conducting an initial quality assessment of all images before they reach human auditors. This preliminary filtering action prepares the workflow by identifying and flagging only the images that need further human evaluation, thereby reducing the overall workload on auditors while maintaining compliance.
3Measurement precision
If manual review of all images is performed to ensure accurate quality assessment, then measurement precision improves, but productivity decreases
Solution Approach 1:
Instead of performing complete manual review on all images, the system applies partial action by conducting manual verification only on the subset of images with low confidence scores. This partial review approach is sufficient to ensure accuracy for uncertain cases while avoiding the excessive workload of reviewing every image manually, thus maintaining productivity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively utilizes automated quality assessment to identify images requiring manual review, improving the efficiency and accuracy of the auditing process without sacrificing regulatory compliance.
Implementation Method 1
filtering the mammogram image using an edge detection filter to generate an edge filtered image
Data Source
AI summary
An apparatus (10) for manually auditing a set (30) of images having quality ratings (38) for an image quality metric assigned to the respective images of the set of images by an automatic quality assessment process (40) includes at least one electronic processor (20) programmed to: generate quality rating confidence values (42) indicative of confidence of the quality ratings for the respective images; select a subset (32) of the set of images for manual review based at least on the quality rating confidence values; and provide a user interface (UI) (27) via which only the subset of the set of images is presented and via which manual quality ratings (46) for the image quality metric are received for only the subset of the set of images.