A Visualization Method of Knowledge Graphs for the Computation and Comprehension of Ultrasound Reports

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Radiology Informatics

Background:

• Knowledge graph visualization is crucial for improving medical decision-making and computer-aided analysis in ultrasound diagnostics.
• Current methods for extracting and visualizing knowledge from unstructured ultrasound reports are limited.

Purpose of the Study:

• To propose an intelligent method for analyzing ultrasound reports via knowledge graph visualization.
• To enhance the extraction of clinical concepts and improve the structured representation of ultrasound knowledge.

Main Methods:

• Developed a novel method for accurate key term network extraction from ultrasound report narratives.
• Proposed a knowledge representation framework for structured and intuitive visualization of ultrasound report data.
• Introduced a knowledge graph completion model to address entity sparsity and enhance visualization accuracy.

Main Results:

• Achieved a significantly higher extraction index (η = 2.69) compared to traditional methods (2.12).
• Outperformed state-of-the-art methods with the highest P (0.85), R (0.89), and F1 (0.87) scores across three datasets.
• Demonstrated effective utilization of embedded knowledge for accurate clinical information retrieval.

Conclusions:

• The proposed intelligent method effectively extracts and visualizes knowledge from complex ultrasound reports.
• This approach enhances the accuracy of using ultrasound knowledge for clinical decision support.
• The method offers a significant advancement in leveraging unstructured data within medical imaging reports.