Leveraging artificial intelligence in clinical management of nAMD

At this year’s EURETINA in Hamburg, Germany, there were several engaging presentations during the Symposium Session that took place on the Friday, 2nd September on the important topic of artificial intelligence (AI) tools for fluid monitoring in neovascular age-related macular degeneration (nAMD).

In his presentation, “The fluid tool in clinical routine – Data from FRB! In Zurich”,
Prof. Daniel Barthelmes, director of the Department of Ophthalmology at University Hospital Zurich, Switzerland, discussed a study that was designed to investigate the feasibility of analysing real-practice optical coherence tomography (OCT) imaging data using an established AI method.

The study aimed to determine the impact of subretinal fluid (SRF), intraretinal fluid (IRF) and pigment epithelium detachment (PED) on visual acuity outcomes at 12 and 48 months after starting anti-VEGF treatment.

The patients enrolled onto the study were treatment naïve and had nAMD, an average age of 78.9 years, and had had no prior treatment. A total of 209 eyes were analysed during the trial.

The Vienna Fluid monitor, an AI-based method for detecting and quantifying retinal fluid, was used to analyse OCT volume scans for SRF, IRF and PED from routine clinical practice and the effect of post-loading phase fluid volumes on visual acuity outcomes.

The study determined that the volume of fluids present was correlated with visual acuity outcomes and that the Vienna Fluid monitor is a more accurate method of assessing the volume of fluid present. Prof. Barthelmes explained that it is imperative to control the volume of fluid to improve patient outcomes, and this should be reflected in a change to the treatment paradigm.

He concluded by stating that the AI tool allows for more precise measurements in real time which can be utilised to improve treatment benefits in daily clinical practice. Patients with high IRF volumes after the loading phase are at risk of losing vision during the maintenance phase. Patients with high macular fluid volumes receive more injections compared to those with low volumes. For SRF and PED, this increased number of injections can prevent vision loss, however, for IRF the increase in injections does not compensate for loss in visual acuity.

Multicentre study

Another speaker in the AI Symposium, Dr Gregor S. Reiter, Laboratory for Ophthalmic Image Analysis, Department of Ophthalmology and Optometry, Medical University of Vienna, Vienna, Austria, took to the podium to highlight recent data from a prospective multicentre Phase III clinical trial using AI in nAMD.

The trial enrolled 290 patients with active nAMD with foveal intra- and/or subretinal fluid who were randomised into receiving AI-support or conventional, qualitative fluid assessment. The patients were aged 50 years or older and all had a best-corrected visual acuity better-than or equal-to 1.0 log MAR. In the BCVA-examiner blinded study, patients were randomised to undergo either conventional qualitative fluid assessment or AI-guided fluid assessment and the resulting impact on injection frequency was analysed.

According to Dr Reiter, the results demonstrated that fluid volumes can be precisely identified, localised and measured in real-time using AI. He said that treatment for nAMD is based on fluid volumes but also emphasised the importance of fluid dynamics.

He concluded by describing how automated tools such as these will enable precision medicine based on fluid guidance to enter real-world management of exudative disease, thus improving clinical outcomes whilst also providing savings in terms of healthcare resources.

Real-time monitoring

Towards the end of the Symposium Session, Stela Vujosevic, head of the Medical Retina Service at the University Hospital,“Maggiore della Carità”, Novara, Italy, provided attendees with some practical guidance on handling the RetInSight Fluid Monitor, a Class IIa medical device that can be used to help support clinical decisions in the monitoring of patients with nAMD in Europe.

The fluid monitor is designed to identify, localise and quantify retinal fluid in routine OCT images. It assigns a label to each OCT pixel, denoting the fluid type as IRF, SRF or PED, and provides its location within the central 1-mm and 6-mm macular area. Fluid measurements are provided in nanolitres.

According to Dr Vujosevic, it is important to distinguish between the specific fluid compartments due to the differences in the impact on visual function and the dynamics of fluid resolution whilst patients are receiving different treatment regimens. She explained that the measurements and readouts of fluid activity enable ophthalmologists to monitor disease progression or response to treatment. The monitor is intended to be compatible with all leading OCT systems.

Dr Vujosevic went onto discuss that the AI-based software can be deployed by the OCT operator under the supervision of an eyecare specialist in the clinic environment, producing a concise and highly accurate real-time report containing patient results whilst the patient is on site. Benefits to patients include shorter waiting times for appointments, less inconvenience and lower cost associated with travelling and absence from work resulting from over-treatment, as well as a lowered risk of vision loss caused by under-treatment.

Meanwhile, eyecare specialists benefit from being able to monitor their patients long term and can optimise and personalise treatment regimens, as well as having access to a more efficient workflow and one which enables then to see more patients.