Automated artificial intelligence-based system for clinical follow-up of patients with age-related macular degeneration

Abstract

Purpose: In this study, we investigate the potential of a novel artificial intelligence-based system for autonomous follow-up of patients treated for neovascular age-related macular degeneration (AMD).

Methods: A temporal deep learning model was trained on a data set of 84 489 optical coherence tomography scans from AMD patients to recognize disease activity, and its performance was compared with a published non-temporal model trained on the same data (Acta Ophthalmol, 2021). An autonomous follow-up system was created by augmenting the AI model with deterministic logic to suggest treatment according to the observe-and-plan regimen. To validate the AI-based system, a data set comprising clinical decisions and imaging data from 200 follow-up consultations was collected prospectively. In each case, both the autonomous AI decision and original clinical decision were compared with an expert panel consensus.

Results: The temporal AI model proved superior at detecting disease activity compared with the model without temporal input (area under the curve 0.900 (95% CI 0.894-0.906) and 0.857 (95% CI 0.846-0.867) respectively). The AI-based follow-up system could make an autonomous decision in 73% of the cases, 91.8% of which were in agreement with expert consensus. This was on par with the 87.7% agreement rate between decisions made in the clinic and expert consensus (p = 0.33).

Conclusions: The proposed autonomous follow-up system was shown to be safe and compliant with expert consensus on par with clinical practice. The system could in the future ease the pressure on public ophthalmology services from an increasing number of AMD patients.

Keywords: age-related macular degeneration; anti-vegf; artificial intelligence; follow-up.

Fig. 1

Stages of development and data utilization. This figure provides an overview of how the various sources of data were used during the training and validation processes. Data sets used for training, tuning and validation are shown, each with the following information displayed: number of cases, whether the data were retrospectively or prospectively collected, and labelling information. For the latter, the source of the labelling and what the labels denote is given. The source of the labelling is accompanied by either an equality sign (‘=’) when expert re‐grading is used, or a tilde (‘~’) if labels are derived indirectly from PRN treatment decisions (refer to the Materials section for more information). The figure differentiates between the validation of the AI component alone (red pictogram ‘AI’) and the validation of the entire system (i.e. the AI component together with the deterministic logic represented by the red pictogram ‘Logic’). Finally, the figure states which sources the current performance metrics were compared with. For a more detailed description of the data sets, see Table 1.

Fig. 2

AI system component overview. This figure provides an overview of the input and output of the AI‐based system, along with a simplified schematic of the internal components (the AI model and the deterministic logic). On the left, inputs are shown: clinical data from the current and previous visit, along with treatment history of the prior 6 months, are directly processed by the deterministic logic component; OCT scans from the previous and current visit are input first into the AI model (along with the fundus photograph for the model that supports it), and then the activity score is passed into the deterministic logic component. After processing, the deterministic logic will output either an autonomous treatment decision (number of injections with a given interval or follow without treatment) or a request for second opinion (elaborated by a reason or a suggestion for further actions; list of possibilities shown on the right in yellow). For a more comprehensive overview of the deterministic logic component, see Fig. S1. T – treatment, IOP – intraocular pressure, BCVA – best‐corrected visual acuity, OCT – optic coherence tomography scan.