lawrence jenga
November 22, 2024 16:35
NVIDIA’s research shows that its AI technology, Eye-AD, detects early signs of Alzheimer’s disease by analyzing retinal scans and provides a non-invasive and rapid screening method.
A groundbreaking development is being made where artificial intelligence (AI) is being used to detect early signs of Alzheimer’s disease through retinal scans. The study, reported by NVIDIA, introduces a deep learning framework called Eye-AD that analyzes high-resolution retinal images to identify subtle changes in the ocular vascular bed often associated with dementia. This innovative approach provides a rapid, non-invasive screening method that can significantly improve the early detection and treatment of cognitive decline.
The importance of early detection
Alzheimer’s disease (AD) currently affects more than 50 million people worldwide, and the number is expected to increase as the global population ages. Early detection is important to improve patient outcomes and quality of life, slow disease progression through timely clinical intervention, and allow families to plan for long-term treatment and support.
The Retina: A Window to the Brain
The retina, often referred to as the ‘window to the brain’, shares an embryonic origin with the brain. Studies have shown that changes in the microvasculature (small blood vessels) of the retina are often associated with cognitive decline. Traditional detection methods, such as MRI and spinal fluid analysis, are costly and invasive, making Eye-AD’s non-invasive approach particularly promising.
Technological advancements in Eye-AD
Eye-AD, developed by researchers, combines convolutional neural networks (CNN) and graph neural networks (GNN) to extract features from retinal images and analyze relationships within and between retinal layers. Use Optical Coherence Tomography Angiography (OCTA) images to visualize blood flow and vascular details and identify clinical biomarkers to predict early-onset Alzheimer’s disease (EOAD) and mild cognitive impairment (MCI).
The model was trained on 5,751 OCTA images from 1,671 patients using PyTorch on a workstation equipped with four NVIDIA GeForce RTX 3090 GPUs, which significantly accelerated training time and processing efficiency for high-resolution images.
Performance and future prospects
Eye-AD demonstrated better accuracy than other models in detecting EOAD, with an Area Under the Curve (AUC) of 0.9355 in the internal dataset and 0.9007 in the external dataset. MCI detection performance was slightly lower, but still achieved an AUC of 0.8630 internally and 0.8037 externally. This study highlights that the deep vascular complex in the retina is a key biomarker for predicting early disease.
The researchers emphasize that Eye-AD represents a substantial advance in dementia detection with the potential for widespread use in cognitive health assessment. Future efforts will focus on validating the model across diverse populations and integrating it with other diagnostic tools to assist physicians in clinical practice.
Eye-AD source code is accessible on GitHub, providing opportunities for further development and research in this promising field.
Read the full study on nature for more insight.
Image source: Shutterstock