Burnt-in Text Recognition from Medical Imaging Modalities - Existing Machine Learning Practices - Vol 2

Aug 5, 2026

Upload by Aurora

Medical imaging modalities store crucial clinical information as burnt-in textual data within the pixel content, including patient demographics, often on ultrasound and X-ray. Recognition of this embedded text supports automated access to image-associated information for medical informatics, aiding clinical decision-making, patient engagement, and treatment recommendations. The review summarizes the significance of burnt-in text recognition, highlights machine-learning methods, and analyzes challenges driven by low resolution and background interference, concluding with recommendations such as advanced deep-learning ensemble approaches.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

Motivation and objectives

OCR and related image-processing workflows

Textual Data Recognition from MIM

Significance for clinical informatics

Value in decision-making and EHR integration

Background literature (ML-based techniques)

Challenges and open issues

Conclusion and recommendations

Osagie, Efosa ORCID logoORCID: [https://orcid.org/0009-0004-3462-7175](https://orcid.org/0009-0004-3462-7175) , Ji, Wei and Helian, Na (2024) Burnt-in Text Recognition from Medical Imaging Modalities: Existing Machine Learning Practices. Journal of Advanced Computational Intelligence and Intelligent Informatics. Downloaded from: [https://ray.yorksj.ac.uk/id/eprint/12902/](https://ray.yorksj.ac.uk/id/eprint/12902/) The version presented here may differ from the published version or version of record. If you intend to cite from the work you are advised to consult the publisher's version: [https://doi.org/10.20965/jaciii.2024.p0103](https://doi.org/10.20965/jaciii.2024.p0103) Research at York St John (RaY) is an institutional repository. It supports the principles of open access by making the research outputs of the University available in digital form. Copyright of the items stored in RaY reside with the authors and/or other copyright owners. Users may access full text items free of charge, and may download a copy for private study or non-commercial research. For further reuse terms, see licence terms governing individual outputs. Institutional Repository Policy Statement RaY Research at the University of York St John For more information please contact RaY at [ray@yorksj.ac.uk](ray@yorksj.ac.uk) [https://doi.org/10.20965/jaciii.2024.p0103](https://doi.org/10.20965/jaciii.2024.p0103) Review: Burnt-in Text Recognition from Medical Imaging Modalities: Existing Machine Learning Practices Efosa Osagiey, Wei Ji, and Na Helian Department of Computer Science, University of Hertfordshire College Lane Campus, Hatfield, Hertfordshire AL10 9AB, United Kingdom E-mail: {e.osagie, w.1.ji, [n.helian}@herts.ac.uk](n.helian}@herts.ac.uk) yCorresponding author [Received February 2, 2023; accepted August 17, 2023] In recent times, medical imaging has become a significant component of clinical diagnosis and examinations to detect and evaluate various medical conditions. The interpretation of these medical examinations and the patient’s demographics are usually textual data, which is burned in on the pixel content of medical imaging modalities (MIM). Example of these MIM includes ultrasound and X-ray imaging. As artificial intelligence advances for medical applications, there is a high demand for the accessibility of these burned-in textual data for various needs. This article aims to review the significance of burned-in textual data recognition in MIM and recent research regarding the machine learning approach, challenges, and open issues for further investigation on this application. The review describes the significant problems in this study area as low resolution and background interference of textual data. Finally, the review suggests applying more advanced deep learning ensemble algorithms as possible solutions. Keywords: medical image character recognition, OCR challenges, burned-in text, medical imaging, medical image processing 1 . Introduction Medical imaging modalities (MIM) are used to probe the human body parts for clinical examinations, and the interpretation of the resulting examinations is burned into the images in the form of textual data, including the patient’s demographics. Still, these textual data come with various constraints due to the unique nature of these images. These include background interference, low contrast, distortion, and low resolution of MIM. The optical character recognition (OCR) technique has been applied to enhance visual interpretation and allow automated or semi-automated text recognition and extraction from MIM [1] . In combination with OCR, sophisticated and extensive image processing techniques are also used to eliminate manual textual data recognition [2] . Most processes involve characterisation, multiple image transformations, and other methods to extract the textual data of interest from MIM [1] . Therefore, an investigation of the problem of low resolution

FAQs

Why is burnt-in text recognition important in medical imaging modalities?

Burnt-in textual data embedded in images (e.g., patient demographics) can be leveraged to support clinical decision-making, patient engagement, and treatment recommendations using ML and related statistical techniques.

What main technical challenges affect recognition of burnt-in text in MIM?

Low resolution and background interference are emphasized as significant problems, along with constraints such as low contrast, distortion, and other image-specific limitations.

What solution direction does the review suggest for future work?

The review suggests applying more advanced deep learning ensemble algorithms as possible solutions to improve recognition performance under challenging imaging conditions.

Aug 5, 2026

Upload by Aurora