An Ensemble Learning Artificial Intelligence Model for Alzheimer's Disease Detection Using OCT

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) diagnosis relies on clinical assessment and costly biomarkers.
• Retinal imaging offers a non-invasive window into neurodegenerative changes associated with AD.
• Deep learning (DL) models show potential for analyzing complex imaging data.

Purpose of the Study:

• To develop and validate an ensemble DL model using Optical Coherence Tomography (OCT) for detecting AD-dementia and early AD.
• To integrate multiple OCT-derived inputs for enhanced diagnostic performance.
• To assess the model's ability to classify mild cognitive impairment (MCI) and AD with PET-confirmed biomarkers.

Main Methods:

• A retrospective case-control study involving participants with AD-dementia, MCI, and cognitively normal controls.
• Development of two base DL models (ONH and macula) using various OCT imaging data.
• An ensemble model was created by integrating the base models for unified classification.
• External validation was performed using independent cohorts with PET-confirmed amyloid-beta status.

Main Results:

• The ensemble model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.943 for detecting AD-dementia in internal validation.
• For external validation, the model showed AUROCs of 0.786 and 0.795 for AD-dementia detection.
• The model achieved AUROCs around 0.79 for detecting AD-MCI (PET-based) in external cohorts.

Conclusions:

• The proposed ensemble DL model effectively detects AD-dementia and early AD using OCT imaging.
• Integrating multiple DL models and OCT inputs enhances diagnostic accuracy.
• This approach enables opportunistic AD screening during routine ophthalmic examinations.