June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Automated Image Enhancement of Near-Infrared Retinal Images
Author Affiliations & Notes
  • Cameron McGlone
    Ophthalmology, University at Buffalo Jacobs School of Medicine and Biomedical Sciences, Buffalo, New York, United States
  • Abhiniti Mittal
    Ophthalmology, University at Buffalo Jacobs School of Medicine and Biomedical Sciences, Buffalo, New York, United States
  • Brian Madow
    Ophthalmology, University at Buffalo Jacobs School of Medicine and Biomedical Sciences, Buffalo, New York, United States
  • Footnotes
    Commercial Relationships   Cameron McGlone None; Abhiniti Mittal None; Brian Madow None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 2385. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Cameron McGlone, Abhiniti Mittal, Brian Madow; Automated Image Enhancement of Near-Infrared Retinal Images. Invest. Ophthalmol. Vis. Sci. 2023;64(8):2385.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : Native near-infrared retinal images obtained with Spectralis OCT often lack sufficient contrast and sharpness necessary to easily identify features such as microaneurysms, dot blot hemorrhages, and vascular markings in patients with diabetic retinopathy (DR). We sought to develop and validate automated software to reliably enhance the images.

Methods : The software was developed on Microsoft Windows 11-based computer utilizing functions available via MATLAB. Multiple algorithms have been explored and tested to achieve optimal output, including brightness, contrasts, sharpness, and histogram optimization in proportional combinations. Each approach was tested on 100 standardized images and the resulting images were evaluated independently by 3 clinicians. A graphical user interface (GUI) was created for facile interaction with the software on Windows platforms. The software allows for the original and enhanced images to be displayed next to each other after selecting the input image. The enhanced images can be archived via a dialog box or automatically to a predetermined file location as lossless PNG files.

Results : Software performance was tested by comparing non-enhanced to enhanced images on a monitor under standardized conditions until maximum subjective results were achieved. The algorithms have been adjusted to enhance parts of the image selectively to avoid oversaturating brighter or darker areas in comparison to manual adjustment. The image processing duration was less than 1.00 seconds per image. The same set of images was tested 3 times and the reliability and reproducibility were found to be 100% based on the evaluation performed by 3 experienced clinicians graders. Additionally, the automated approach was found to be superior to the non-automated approach. Microaneurysms and small hemorrhages were detected at a higher rate on the enhanced images than during clinical examination.

Conclusions : A reliable and highly reproducible image enhancement software for near-infrared retinal images has been developed and validated. The facile GUI allows for image input selection however the image processing is completely operator independent. The enhanced images offer higher image contrast and definition, allowing for better discrimination of retinal features. The software can enhance physician tools for the diagnosis and management of patients with retinal conditions and could be deployed for purposes of telemedicine and DR screening.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×