Skip to main content
  • News in Review

    Using AI to Help Diagnose Optic Neuropathies

    Add to My Bookmarks

    Download PDF

    Researchers from the Brain and Optic Nerve Study With Artificial In­telligence (BONSAI) Consortium and the Singapore Eye Research Institute developed a deep learning (DL) system to automatically identify high-quality images of the optic nerve head (ONH). The system accurately distinguished between poor-and high-quality fundus photographs and may help improve the selection of optimal-quality photographs for diagnosis of optic neuropathies.1

    “In contrast to our initial assumption that the presence of pathologically blurred discs in otherwise high-qual­ity images would result in interpretation errors and image misclassification as of poor quality, the system was able to accurately determine the quality of retinal images,” said Dan Milea, MD, PhD, at the Singapore Eye Research Institute.

    Graph of AI results with optic nerve head images.

    AI RESULTS. Performance of the DL system in a multiethnic external-testing dataset (AUC = area under the curve, ROC = receiver operating characteristic).

    Study rationale. Although DL algorithms have been developed to automate the identification of low-quality retinal images with blurred areas used for the diagnosis of diabetic retinopathy or glaucoma, papilledema is “by definition a visually blurred area due to optic disc swelling,” Dr. Milea noted. As a result, the team developed an algorithm to evaluate the quality of retinal images irrespective of the presence of blurred discs.

    Study specifics. The BONSAI re­searchers developed, trained, and tested their DL system on “5,015 mydriatic and nonmydriatic ocular photographs from 31 neuro-ophthalmology centers in 20 countries,” Dr. Milea said. Disease conditions represented included papilledema and glaucoma; images of eyes without pathology also were included. Three experts independently deter­mined the quality of images as good, borderline, or poor.

    For high-quality images, the BON­SAI system provided 91.4% (95% CI, 90%-92.9%) accuracy, 93.8% (95% CI, 92.5%-95.2%) sensitivity, and 75.9% (95% CI, 69.7%-82.1%) specificity. High performance was also obtained for fundus photographs of poor and borderline quality. The system’s overall accuracy in distinguishing between the three levels of image quality was 90.6% (95% CI, 89.1%­92.1%).

    Potential clinical utility. This DL algorithm could be connected to a fundus camera for real­time quality assessment of an image, Dr. Milea said. “Fundus photographs of high quality could be used to detect papilledema or other optic disc abnormalities, whereas those of poor quality would be marked for reacquisition.”

    —Christos Evangelou, PhD

    ___________________________

    1 Chan E et al. Diagnostics. 2023;13(1):160.

    ___________________________

    Relevant financial disclosures: Dr. Milea—None.

    For full disclosures and the disclosure key, see below.

    Full Financial Disclosures

    Dr. Harasymowycz AbbVie: C; Alcon: C; Glaukos: C; New World Medical: C; Nova Eye: C.

    Dr. Milea Optomed: C.

    Dr. Sahay Enterx Bio: EO; NEI: S

    Dr. Sodhi HIF Therapeutics: PS; NIH: S; Research to Prevent Blindness: S; TEDCO: S.

    Disclosure Category

    Code

    Description
    Consultant/Advisor C Consultant fee, paid advisory boards, or fees for attending a meeting.
    Employee E Hired to work for compensation or received a W2 from a company.
    Employee, executive role EE Hired to work in an executive role for compensation or received a W2 from a company.
    Owner of company EO Ownership or controlling interest in a company, other than stock.
    Independent contractor I Contracted work, including contracted research.
    Lecture fees/Speakers bureau L Lecture fees or honoraria, travel fees or reimbursements when speaking at the invitation of a commercial company.
    Patents/Royalty P Beneficiary of patents and/or royalties for intellectual property.
    Equity/Stock/Stock options holder, private corporation PS Equity ownership, stock and/or stock options in privately owned firms, excluding mutual funds.
    Grant support S Grant support or other financial support from all sources, including research support from government agencies (e.g., NIH), foundations, device manufacturers, and\or pharmaceutical companies. Research funding should be disclosed by the principal or named investigator even if your institution receives the grant and manages the funds.
    Stock options, public or private corporation SO Stock options in a public or private company.
    Equity/Stock holder, public corporation US Equity ownership or stock in publicly traded firms, excluding mutual funds (listed on the stock exchange).

     

    More from this month’s News in Review