IBM Watson being trained to identify abnormalities in retina images

IBM Research announced Tuesday that researchers have trained a research version of Watson to recognise abnormalities in retina images. The company noted that the advancement could "offer doctors greater insights and speed in their early identification of patients who may be at risk of eye diseases, such as glaucoma."

Joanna Batstone, lab director at IBM Research Australia, remarked that "medical image analysis with cognitive technology has the capacity to fundamentally change the delivery of healthcare services," adding that "cognitive technology holds immense promise for confirming the accuracy, reproducibility and efficiency of clinicians' analyses during the diagnostic workflow."

The company noted that while the research began in 2015, the latest work focused on streamlining some of the physicians' manual processes, such as evaluating the quality of retina scans and ranking possible indicators of glaucoma.  

Specifically, the researchers applied deep learning techniques and image analytics technology to 88 000 de-identified retina images, showing Watson's ability to measure the "ratio of the optic cup to disc, which is a key sign of glaucoma, with statistical performance as high as 95 percent," IBM said. Further, the technology was trained to distinguish between left and right eye images with up to 94-percent confidence. IBM added that the technology is expected to expand "to detect features of other eye diseases such as diabetic retinopathy and age-related macular degeneration."

Last June, IBM Research announced plans to undertake research with Melanoma Institute Australia to help further advance the identification of melanoma using cognitive technology. IBM also has a research agreement with Molemap, which uses advanced visual analytics to examine more than 40 000 data sets including images and text.