Urgency-Based Worklist Prioritization for Radiology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current workflow prioritization systems for radiological image studies, such as FIFO and classification-based methods, fail to consider the severity of conditions and prioritize within or between similarly severe cases, leading to inefficient distribution and review of image studies.
Innovation Solution
A computer-implemented method using a deep learning network trained with previously read image studies to predict an urgency score for unread image studies, incorporating patient-specific and radiologist-specific data to generate a prioritized worklist.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If FIFO method is used for prioritization, then the system is simple to implement, but it does not consider severity of conditions leading to inefficient workflow
Solution Approach 1:
The patent replaces the simple mechanical FIFO queue system with an intelligent prioritization system that uses machine learning models to analyze clinical data, image features, and workflow parameters. This substitution enables the system to automatically assess urgency and severity without manual intervention, resolving the contradiction between system simplicity and workflow efficiency.
Solution Approach 2:
The invention introduces multiple dynamic parameters including clinical urgency indicators, image-based severity features, radiologist expertise levels, and workflow constraints. These parameters are continuously evaluated by the prioritization algorithm to dynamically adjust case priority, transforming the static FIFO approach into a flexible, efficiency-optimized system.
2Reliability
If classification-based prioritization is used, then critical conditions can be identified, but prioritization within or between similarly severe conditions is not achieved
Solution Approach 1:
The patent segments the prioritization process into multiple hierarchical levels: first classifying by broad condition categories (critical, urgent, routine), then further segmenting within each category using detailed image features and clinical parameters. This multi-level segmentation enables precise differentiation between similarly severe conditions, improving both reliability and prioritization precision.
Solution Approach 2:
The invention adds multiple dimensions to the prioritization framework beyond simple classification, including temporal urgency, spatial distribution across radiologists, expertise matching dimensions, and workflow load balancing. These additional dimensions enable nuanced prioritization within and between classification groups, resolving the limitation of traditional classification-based systems.
3Speed
If image studies are distributed to different departments and countries, then reading speed increases, but coordination and prioritization become more complex
Solution Approach 1:
The patent implements a universal prioritization platform that serves multiple functions: case triage, radiologist matching, workload balancing, and performance monitoring. This multi-functional system handles distribution across departments and countries through a single coordinated interface, reducing complexity while maintaining high reading speeds through efficient resource allocation.
Solution Approach 2:
The invention introduces an intelligent intermediary prioritization system that acts as a mediator between image acquisition, multiple radiology departments, and final reporting. This intermediary layer standardizes case distribution, manages cross-departmental coordination, and maintains centralized prioritization control, simplifying the complex multi-site workflow while preserving reading speed benefits.
Data Source
AI summary
A system and method for training a deep learning network with previously read image studies to provide a prioritized worklist of unread image studies. The method includes collecting training data including a plurality of previously read image studies, each of the previously read image studies including a classification of findings and radiologist-specific data. The method includes training the deep learning neural network with the training data to predict an urgency score for reading of an unread image study.