Generative Knowledge Graphs for Cross-Domain Medical Imaging Queries

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing knowledge graph construction techniques for medical imaging scanners are domain-specific and lack generalizability across different technical domains, requiring extensive retraining to adapt to new domains.

Innovation Solution

Employ generative text-to-text neural networks to construct knowledge graphs by iteratively executing on a design discovery tree associated with the medical imaging scanner, using a generative text-to-text model to synthesize textual content and build a knowledge graph representing electronic documents, and convert natural language queries to structured queries for answering.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If domain-specific techniques are used to construct knowledge graphs, then the knowledge graph can be built for a specific domain, but the technique cannot be easily implemented across different domains

Engineering Contradiction:
Improvedomain generalizabilityVSAvoidretraining requirement
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies universality by transforming domain-specific knowledge graph construction techniques into a universal framework that can operate across multiple domains. The system uses a domain-agnostic neural network architecture that processes electronic documents through standardized pipelines (parsing, entity recognition, relation extraction) regardless of the specific domain, enabling the same system to construct knowledge graphs for medical imaging scanners, consumer electronics, or other technical domains without domain-specific customization

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent applies parameter changes by modifying the input representation and processing parameters rather than the fundamental architecture. The system accepts electronic documents in standardized formats and uses configurable parameters such as entity types, relation types, and extraction thresholds that can be adjusted for different domains without changing the core system structure, thereby achieving domain adaptability through parameter adjustment rather than architectural redesign

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If extensive retraining is performed to adapt to new domains, then the knowledge graph construction can be customized for specific domains, but the implementation time and resources increase significantly

Engineering Contradiction:
Improvedomain customizationVSAvoidretraining time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-configuring the knowledge graph construction system with universal schemas, standardized entity and relation types, and pre-trained language understanding capabilities. The neural network is pre-trained on general technical documentation in a self-supervised manner before deployment, so that when deployed to a new domain, it requires only minimal domain-specific parameter adjustments rather than extensive retraining, thereby reducing implementation time

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies copying by using template-based knowledge graph schemas and reusable processing pipelines that can be copied and deployed across different domains. The system maintains domain-agnostic templates for entities, relations, and graph structures that can be instantiated for any domain by simply providing domain-specific vocabulary and relationship definitions, eliminating the need to rebuild the entire knowledge graph construction system for each domain

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4592896A1Knowledge graph construction via generative artificial intelligence
Publication Date: 2025.07.30 GE PRECISION HEALTHCARE LLC
  • EP4592896A1 patent drawingFigure 1
  • EP4592896A1 patent drawingFigure 2
  • EP4592896A1 patent drawingFigure 3

AI summary

Systems or techniques that facilitate knowledge graph construction via generative artificial intelligence are provided. In various embodiments, a system can access a plurality of electronic documents (e.g., 106) associated with design or fabrication of a medical imaging scanner (e.g., 104). In various aspects, the system can construct a knowledge graph (e.g., 206) representing the plurality of electronic documents, by iteratively executing a generative text-to-text neural network (e.g., 202) on a design discovery tree (e.g., 204) associated with the medical imaging scanner. In various instances, the system can access a natural language query (e.g., 1602) regarding the medical imaging scanner and can convert, via execution of another neural network (e.g., 1604), the natural language query to a structured query (e.g., 1606). In various cases, the system can execute the structured query over the knowledge graph, thereby yielding an electronic answer (e.g., 1608) to the natural language query.