Exam Prefetching via Anatomical Region Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inconsistent tagging of body parts in medical imaging systems leads to inefficiencies and safety risks, particularly in large healthcare institutions with numerous exam types, where manual configuration of links between master files is complex and time-consuming.
Innovation Solution
Automatically categorize medical examinations into body parts by evaluating language in exam descriptions and using image analytics, such as Watson, to reduce the number of required links and enhance system configuration efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual configuration of links between master files is used to categorize body parts, then system configuration can be performed, but it is complex and time-consuming
Solution Approach 1:
The system performs preliminary automatic categorization of examinations into body parts using image analytics and natural language processing before the user needs to access them. Keywords are extracted from exam descriptions and mapped to anatomical regions in advance, so that when users search or prefetch studies, the categorization work is already complete, eliminating manual configuration time.
Solution Approach 2:
The system enables self-service automatic categorization by extracting keywords from examination descriptions and automatically mapping them to body parts without requiring manual intervention. The image analytics engine and natural language processing automatically perform the categorization task that would otherwise require manual configuration of links between master files.
2Adaptability or versatility
If inconsistent tagging of body parts is used in medical imaging systems, then various exam types can be stored, but inefficiencies and safety risks occur
Solution Approach 1:
The system changes the parameter of body part tagging from inconsistent manual labels to standardized anatomical regions derived from image analytics and keyword extraction. By transforming the tagging parameter into a structured format based on actual image content and examination descriptions, the system maintains versatility for various exam types while ensuring consistent and reliable body part identification.
Solution Approach 2:
The system introduces an intermediary layer of image analytics and natural language processing between the raw examination data and the body part categorization. This intermediary automatically extracts meaningful keywords from exam descriptions and maps them to standardized anatomical regions, resolving the inconsistency problem while preserving the ability to handle diverse exam types.
3Speed
If fast prefetching is implemented, then study retrieval speed is improved, but accuracy of relevant study selection may be compromised
Solution Approach 1:
The system performs preliminary organization of examinations into standardized body part categories using image analytics and keyword extraction. This pre-categorization enables fast prefetching by creating an efficient indexing structure, while the accuracy is maintained through the use of multiple keywords per body part and comprehensive keyword extraction from examination descriptions, ensuring that relevant studies are accurately identified during prefetching operations.
Data Source
AI summary
Inference of appropriate anatomical region from inconsistent descriptions in order to provide fast and accurate prefetching is provided. In various embodiments, a first plurality of user-configurable rules is read from a data store. Each rule maps to a user-configurable anatomical region. A plurality of studies is accessed from the image archive. Each of the plurality of studies has associated metadata. The plurality of rules is applied to the metadata associated with the plurality of studies to determine an anatomical region of each of the plurality of studies. Based on the anatomical regions of the plurality of studies and one or more additional rule, a subset of the plurality of studies is selected for display to a user on a display.