Layout-aware information extraction from semi-structured medical images.

Area of Science:

• Medical Informatics
• Computer Vision
• Natural Language Processing

Background:

• Medical images contain valuable patient information within embedded textual data.
• Extracting this structured information is challenging due to text discontinuity and Optical Character Recognition (OCR) errors.

Purpose of the Study:

• To develop a method for extracting structured textual information from semi-structured medical images.
• To address the challenges posed by spatial discontinuity and OCR inaccuracies in medical image text extraction.

Main Methods:

• Propose a domain-specific language (ODL) for describing text value and layout within images.
• Develop an ODL parser that uses value and spatial constraints to associate image text with a data structure.
• Evaluate the ODL parser on a dataset of real medical images.

Main Results:

• The ODL parser demonstrates superior extraction accuracy compared to existing methods.
• The system shows improved tolerance for OCR inaccuracies and positional variations in text.
• Accuracy can be further enhanced through learning from manual corrections.

Conclusions:

• ODL provides an effective solution for structured information extraction from medical images.
• The proposed parser robustly handles common challenges in medical image text analysis.
• The approach offers a pathway for more accurate and reliable medical data extraction.