Multi-omic analysis of deep learning-derived phenotypes links ophthalmic imaging to cardiovascular and neurological traits

2026-06-16 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Artificial Intelligence & Machine Learning, Health & Medical Research, Life Sciences & Biology · Depth: Expert, extended

Summary

A multi-omic analysis pipeline, integrating physiological, radiomic, metabolomic, and genomic data from the UK Biobank, has linked deep learning-derived ophthalmic imaging features to a range of cardiovascular and neurodegenerative diseases. Researchers trained retinal adversarial autoencoders (Ret-AAE) on 63,946 color fundus photographs (CFP) and 88,972 optical coherence tomography (OCT) images, generating 256-dimensional embeddings. These embeddings were significantly associated with conditions like ischemic heart disease, cerebrovascular disease, hypertension, heart failure, stroke, Parkinson's disease, and various dementias, including Alzheimer's. The study found that CFP-derived embeddings had more cardiovascular associations, while OCT-derived embeddings showed more neurological trait associations. This work provides converging evidence that ophthalmic features reflect complex, multisystem biological processes, connecting them to lipid metabolism and specific gene sets related to neurodegenerative pathology.

Key takeaway

For AI Scientists and Machine Learning Engineers developing diagnostic tools, you should explore multi-omic integration with deep learning-derived ophthalmic phenotypes. Your models can identify early indicators for cardiovascular and neurodegenerative diseases, such as heart failure or Parkinson's. Consider utilizing both OCT and CFP data, as they offer distinct, complementary insights into systemic health, guiding the development of more explainable and personalized medicine approaches.

Key insights

Deep learning-derived ophthalmic imaging features serve as multi-omic biomarkers for systemic cardiovascular and neurodegenerative disease risk.

Principles

• Ophthalmic features indicate multisystem biological processes.
• OCT and CFP provide complementary systemic health insights.
• Deep learning extracts complex, abstract imaging biomarkers.

Method

A multi-omic pipeline trained Ret-AAE models on OCT and CFP images to generate 256-dimensional embeddings. These were analyzed against physiological, radiomic, metabolomic, and genomic data using various statistical methods.

In practice

• Apply deep learning to extract subtle ophthalmic imaging biomarkers.
• Integrate multi-omic data for holistic disease risk assessment.
• Differentiate CFP and OCT for specific cardiovascular/neurological predictions.

Topics

• Multi-omic Analysis
• Deep Learning
• Ophthalmic Imaging
• Cardiovascular Disease
• Neurodegenerative Disease
• Retinal Biomarkers

Code references

• TomJulian/Ret-AAE

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• EyeMVP: OCT-Informed Fundus Representation Learning via Paired CFP--OCT Pretraining — EyeMVP uses paired CFP-OCT pretraining to enrich CFP representations with OCT depth information for improved retinal analysis.
• Identification of shared oxidative stress related hub genes in NAFLD and atherosclerosis using bioinformatics and machine learning — Bioinformatics and machine learning can identify shared oxidative stress-related hub genes in NAFLD and atherosclerosis.
• REVEAL: Multimodal Vision-Language Alignment of Retinal Morphometry and Clinical Risks for Incident AD and Dementia Prediction — Jointly modeling retinal images and clinical risk factors via VLM improves early AD and dementia prediction.
• Three-Dimensional Retinal Microvasculature Restoration in OCT Angiography — A deep learning algorithm accurately restores 3D retinal microvasculature from single OCTA volumes, overcoming imaging artifacts.
• A three-dimensional multi-modal foundation model for optical coherence tomography — OCTCube-M integrates 3D OCT with other retinal imaging modalities for enhanced diagnostic and prognostic capabilities.
• Beyond Independent Frames: Latent Attention Masked Autoencoders for Multi-View Echocardiography — LAMAE uses latent attention to integrate multi-view echocardiography data, improving cardiac representation and transferability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.