Photoreceptor Disease at Ambiguous Transition Zones in Inherited Retinal Degenerations

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Inherited retinal degenerations (IRDs) impact photoreceptor outer segments (OS).
• Optical coherence tomography (OCT) measures the inner and outer segment junction (IS/OS) or ellipsoid zone (EZ) for IRD outcomes.
• The EZ termination point can be ambiguous, complicating disease assessment.

Purpose of the Study:

• To characterize outer segment (OS) disease at each retinal location using widefield OCT.
• To develop an AI model for classifying OS disease states.
• To identify predictors of disease progression in IRDs.

Main Methods:

• Widefield OCT imaging was performed on 17 IRD patients (2 ABCA4-associated, 15 RHO-associated) over ~2 years.
• A two-level artificial intelligence architecture classified outer retinal disease into definitely present OS (DPOS), definitely absent OS (DAOS), or questionably absent OS (QAOS).
• OS maps were compared with serial imaging to assess stability and progression.

Main Results:

• The AI model achieved 95% accuracy in classifying OS disease states.
• Wider questionably absent OS (QAOS) regions indicated a more extended transition from disease to health.
• Progression was predicted by proximity to more diseased areas; DPOS and QAOS pixels near disease showed a 10-fold higher progression likelihood.

Conclusions:

• The assumption of abrupt EZ transitions in IRD outcomes may not align with disease pathophysiology.
• Ambiguous transition zones and their neighborhoods are key indicators of high progression risk.
• These regions can serve as efficient endpoints for clinical trials in IRDs.