A novel interpretable deep learning model, named Eye-AD, has been developed to identify early-onset Alzheimer's disease (EOAD) and mild cognitive impairment (MCI) using optical coherence tomography angiography (OCTA) images.

Recent advancements in artificial intelligence (AI) and deep learning have significantly improved the analysis of ocular imaging for Alzheimer's detection.

Eye-AD has demonstrated superior performance compared to existing state-of-the-art methods in both internal and external datasets for detecting EOAD and MCI, achieving an area under the curve (AUC) of 0.9355 for EOAD detection.

The model's findings indicate that retinal changes associated with EOAD and MCI primarily affect the deep vascular complex (DVC), suggesting it as a sensitive biomarker for early detection.

Validation through three public datasets showed high Dice coefficients for optic disc and optic cup segmentation, indicating the model's accuracy.

The method significantly reduces computational complexity and inference time, achieving processing speeds of approximately 24 milliseconds per image.

Eye-AD leverages the unique characteristics of OCTA data to explore relationships within retinal layers, enhancing detection accuracy.

The study emphasizes the potential of OCTA imaging for diagnosing dementia more efficiently compared to traditional methods.

Clinical studies have shown significant changes in retinal vasculature in Alzheimer's patients, which can be detected using ophthalmic imaging techniques.

The retina serves as an accessible window to study early neurodegeneration and microvascular changes related to dementia.

Eye-AD provides valuable insights into the effects of dementia on retinal structures and can facilitate rapid screening in large populations.

At the 128th Annual American Academy of Ophthalmology Meeting, new data on AI-based home optical coherence tomography (OCT) was presented, highlighting its comparison to standard in-office OCT scans.