A serialization method for digitizing the image-based medical laboratory report.

Area of Science:

• Medical Informatics
• Computer Vision
• Digital Health

Background:

• Medical laboratory reports are often submitted as mobile phone images for teleconsultations.
• Digitizing these reports is crucial for efficient data management and accessibility.
• Existing Optical Character Recognition (OCR) technologies lack a specialized, serialized process for medical report image digitization.

Purpose of the Study:

• To develop and validate a serialization method for digitizing image-based medical laboratory reports.
• To accurately extract and digitally store relevant medical report content.
• To streamline the process of preparing medical reports for teleconsultation applications.

Main Methods:

• Collected and annotated a dataset of 330 image-based medical laboratory reports.
• Trained a layout analysis model using a pre-trained model on the annotated dataset.
• Integrated layout analysis with text detection and recognition models to extract digital content.
• Adjusted the layout of digital content and stored it as a docx file.

Main Results:

• Developed a robust serialization method for digitizing medical laboratory report images.
• Successfully extracted and restored the format of report content.
• Enabled shielding of sensitive and irrelevant information, focusing on key data.
• Achieved accurate digitization of medical laboratory report content.

Conclusions:

• The proposed serialization method effectively digitizes image-based medical laboratory reports.
• This facilitates the correct display of essential medical information for teleconsultations.
• The method successfully removes extraneous data, enhancing data privacy and usability.