Multiple instance learning detects peripheral arterial disease from high-resolution color fundus photography
Simon Mueller1, Maximilian W M Wintergerst1, Peyman Falahat1
1Department of Ophthalmology, University Hospital Bonn, 53127, Bonn, Germany.
Scientific Reports
|January 27, 2022
Summary
Early peripheral arterial disease (PAD) detection may be possible using artificial intelligence on retinal images. A deep learning model analyzed color fundus photography, identifying biomarkers for earlier diagnosis and improved patient monitoring.
Area of Science:
- Ophthalmology
- Cardiology
- Artificial Intelligence
Background:
- Peripheral arterial disease (PAD) stems from atherosclerosis, significantly increasing morbidity and mortality in the elderly.
- Current diagnostic methods often fail to detect early-stage PAD, delaying crucial interventions.
- Effective PAD management relies on addressing risk factors like smoking, hypercholesterolemia, and hypertension.
Purpose of the Study:
- To investigate the potential of a deep learning model to detect atherosclerosis, the underlying cause of PAD, using color fundus photography (CFP).
- To explore the use of a convolutional neural network (CNN) architecture for earlier PAD diagnosis and enhanced disease monitoring.
- To identify retinal imaging biomarkers indicative of PAD through advanced AI analysis.
Main Methods:
- An exploratory study utilizing a deep attention-based Multiple Instance Learning (MIL) architecture on 135 CFP examinations.
- Employing high-resolution CFP images by partitioning them into patches to analyze subtle vascular structure variations.
- Developing a CNN architecture that converts image patches into feature vectors and assigns attention weights to determine their importance.
Main Results:
- The best-performing deep learning model achieved an ROC AUC score of 0.890.
- Visualization of attention weights indicated significant weighting (p < 0.001) of the optic disc and temporal arcades compared to retinal background.
- The findings suggest potential ocular involvement in PAD and highlight the model's ability to identify relevant image regions.
Conclusions:
- Deep learning approaches, specifically attention-based MIL on CFP, show feasibility for detecting PAD.
- Retinal imaging biomarkers identified through AI may facilitate earlier PAD diagnosis and improve patient management.
- This AI-driven method offers a promising avenue for non-invasive PAD screening and monitoring.
Related Concept Videos
01:21Peripheral Arterial Disease II: Clinical Manifestations and Diagnostic Evaluation
51
Clinical manifestationsPeripheral Arterial Disease (PAD) manifests through a range of symptoms, from the characteristic intermittent claudication to atypical presentations and severe complications in advanced stages. Intermittent claudication, a hallmark symptom of PAD, presents as exercise-induced muscle pain that typically resolves within minutes of rest. This pain is reproducible and stems from inadequate blood flow, leading to the accumulation of lactic acid produced during anaerobic...
51
01:19Imaging Studies VII: Vascular Imaging
89
DefinitionRenal angiography, also known as renal arteriography, is an imaging technique used to obtain a comprehensive view of blood flow and the vascular structure of blood vessels in the kidneys and surrounding areas.PurposeRenal angiography detects blood vessel abnormalities in the kidneys, such as aneurysms, stenosis, thrombosis, vascular tumors, and renal artery stenosis. It evaluates kidney function and guides interventional treatments like angioplasty or stent placement.Pre-Procedure...
89