A digital second opinion for AMD

A digital diagnostician is proving successful at picking whether a patient has wet age-related macular degeneration (AMD), according to studies by the developer.

Massachusetts-based not-for-profit company Draper is designing an algorithm to look at optical coherence tomography (OCT) images, to distinguish which patients might benefit from wet AMD treatments.

Draper’s chief data scientist, Dr John Irvine, told OT that: “Today, the clinician makes an informed, but ultimately subjective, assessment of the patient’s condition and expected response to treatment. Any information that can assist in that decision process is helpful.”

The algorithm was put through its paces using the OCT images of 75 patients with wet, dry or no AMD in a recent study presented at the Association for Research in Vision and Ophthalmology conference (1–5 May, Seattle).

The algorithm categorised each image and, for the AMD patients, its conclusions were compared to how well the patients’ eyes responded to the treatment available for the wet form of the disease.

Dr Irvine explained: “As it currently stands, our method predicts the response to therapy correctly 85% of the time, so this should be useful in practice. We believe that through analysis of a larger and more diverse set of data we can improve the performance of our methods.”

He believed, after the system was enhanced, the tool would prove useful to both UK optometrists and ophthalmologists in future.

“Our goal is to provide a tool that operates with the OCT imagery that can be collected routinely and provide useful information in a matter of seconds,” he added.

Draper is at the moment using existing AMD patient data to further refine the algorithm and is hoping to perform a study using 200 patients being currently treated for AMD as well, Dr Irvine said.

On the question of clinicians being able to correctly diagnose wet AMD from the dry form of the disease, the company said it did not have information on such success rates.

However, Dr Irvine concluded: “Our method is objective and repeatable. This means it can be used as a measure of performance for developing and evaluating new treatments.”