Medical Report Text Structuring via Graph Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical reports are often unstructured, making them not machine-readable, non-standardized, and difficult to analyze, with conventional methods for creating structured reports being inflexible and not fully effective in capturing medical nuances.
Innovation Solution
A method that involves parsing medical text to extract data elements, annotating them using medical knowledge bases, establishing relationships, and embedding the annotated data elements into target structures to generate a structured text, allowing for flexible and standardized reporting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If free-text reports are used, then physicians have high freedom to express personal preferences and capture medical nuances, but the reports become not machine-readable, non-standardized, and difficult to analyze
Solution Approach 1:
The free-text report is segmented into structured components by identifying and extracting key entities (findings, impressions, diagnoses) and their relationships. The system divides the unstructured text into organized elements that can be mapped to a target structure, enabling both preservation of physician expression freedom and creation of machine-readable structured output.
Solution Approach 2:
A graph structure serves as an intermediary representation between the free-text input and the target structured format. The graph captures entities, attributes, and relationships in a standardized way that bridges the gap between flexible free-text expression and rigid structured requirements, allowing transformation to multiple target structures.
2Extent of automation
If pre-generated report templates are used, then reports become machine-readable and standardized, but the approach is not flexible enough to adapt to peculiarities of a given medical case
Solution Approach 1:
The system uses dynamic structure selection where the target structure is not fixed but chosen based on the specific medical case characteristics. The graph representation allows flexible adaptation to different case types while maintaining standardization, as the same graph can be embedded into different target structures depending on the case requirements.
Solution Approach 2:
The system applies different structural qualities to different parts of the report based on local case requirements. Rather than forcing a single global template, the graph embedding approach allows local adaptation of the target structure to match the peculiarities of each medical case while maintaining overall standardization.
3Productivity
If speech recognition with predefined keywords is used, then reports are created from dictated audio files, but only predefined section headings are inserted without providing high degree of structure
Solution Approach 1:
The system performs preliminary parsing and graph construction from the dictated text before final structure embedding. By pre-identifying entities, attributes, and relationships in the dictated text and representing them as a graph, the system prepares the structured representation in advance, enabling both efficient processing of dictation and generation of highly structured output.
Data Source
AI summary
Methods, systems, and computer program products for structuring medical report texts are provided. In a method, text is received and parsed to obtain a plurality of data elements, connected by semantic connectors. Based thereon, a graph of annotated data elements is generated. In particular, each of the obtained data elements is annotated using medical knowledge bases, which are indicative of medical vocabulary and/or medical ontology and/or medical statistics. To generate a graph, relationships are established between the annotated data elements. The generated graph is embedded into one of a plurality of target structures, wherein each of the target structures has an order criterion. By the embedding, a structured text is provided, which comprises the annotated data elements ordered according to the order criterion of the one of the plurality of target structures, into which the generated graph has been embedded.