Due to poor air quality, the Academy San Francisco offices will be closed Friday, Nov. 16. 

  • IBM
    Glaucoma, Retina/Vitreous

    Computer researchers from IBM Melbourne say the artificial intelligence system Watson is acing its training to identify eye abnormalities.

    A version of Watson, famous for outscoring human champions at the trivia show Jeopardy!, can now detect signs of glaucoma by accurately measuring the cup to disc ratio with 95% confidence. The system can also distinguish between images of right and left eyes with 94% confidence.

    The system uses deep learning techniques and image analytics technology to review data, score evidence and then present the most likely hypothesis. Researchers fed Watson 88,000 retinal images from EyePACS, stripped of patient identifying data, and allowed it to teach itself how to find differences.

    The team hopes to soon use Watson in a clinical setting to help physicians identify patients at risk of diseases such as glaucoma. By streamlining the laborious manual tasks of reviewing retinal images and distinguishing between left and right scans, Watson could allow ophthalmologists to care for increasing patient populations with greater speed and efficiency.

    “It is estimated that at least 150,000 Australians have undiagnosed glaucoma, with numbers expected to rise due to our rapidly aging population. It is critical that every Australian has access to regular eye examinations throughout their life so that diseases like glaucoma and diabetic retinopathy can be detected and treated as early as possible,” said Dr. Peter van Wijngaarden, principal investigator at the Centre for Eye Research Australia, Department of Ophthalmology, University of Melbourne.  “There is a real need for resources that allow all Australians to access regular eye examinations and the development of image analytics and deep learning technology will provide great promise in this area.”

    IBM Research has 11 other collaborative labs worldwide focused on combining cognitive technology with medical images for analysis of other eye diseases, melanoma, breast cancer, and lung cancer.

    “Medical images represent a rich source of data for clinicians to make early diagnosis and treatment of disease, from assessing the risk of melanomas to identifying eye diseases through the analysis of retinas. Cognitive technology holds immense promise for confirming the accuracy, reproducibility and efficiency of clinicians’ analyses during the diagnostic workflow,” said Dr. Joanna Batstone, vice president and lab director at IBM Research Australia.