Predicting Cardiovascular Risk Factors from Retinal Fundus Photographs using Deep Learning
By Ryan Poplin, A. Varadarajan, Katy Blumer, Yun Liu, M. McConnell, Greg S. Corrado, Lily H. Peng, Dale R. Webster. 2017
1,503 citations recorded by the source corpus.
Traditionally, medical discoveries are made by observing associations and then designing experiments to test these hypotheses. However, observing and quantifying associations in images can be difficult because of the wide variety of features, patterns, colors, values, shapes in real data. In this paper, we use deep learning, a machine learning technique that learns its own features, to discover new knowledge from retinal fundus images. Using models trained on data from 284,335 patients, and validated on two independent datasets of 12,026 and 999 patients, we predict cardiovascular risk factors not previously thought to be present or quantifiable in retinal images, such as such as age (within 3.26 years), gender (0.97 AUC), smoking status (0.71 AUC), HbA1c (within 1.39%), systolic blood pressure (within 11.23mmHg) as well as major adverse cardiac events (0.70 AUC). We further show that our models used distinct aspects of the anatomy to generate each prediction, such as the optic disc or blood vessels, opening avenues of further research.
Subfields: Epidemiology, Ophthalmology, Radiology, Nuclear Medicine and Imaging, Medical Imaging
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Epidemiology
- Irene Y. Chen
- Greg S. Corrado
- Arvid Lundervold
- Peter F. Schulam
- Volodymyr Kuleshov
- Marzyeh Ghassemi
- Praneeth Vepakomma
- Tristan Naumann
In Ophthalmology
In Radiology, Nuclear Medicine and Imaging
- Praneeth Vepakomma
- Arvid Lundervold
- Ramesh Raskar
- Abhishek Singh
- Otkrist Gupta
- B. van Ginneken
- Hiroshi Fujita
- Faisal Mahmood
In Medical Imaging
- Sébastien Ourselin
- Arvid Lundervold
- Geert J. S. Litjens
- Ming Y. Lu
- Richard J. Chen
- Tiffany Y. Chen
- Drew F. K. Williamson
- Faisal Mahmood
Citations in the corpus
- NoPeek: Information leakage reduction to share activations in distributed deep learning (2020, background): “Previous research [61] has shown it is possible to predict personal attributes of a person like gender, smoking habits etc.”
- An overview of deep learning in medical imaging focusing on MRI (2018, background): “This is illustrated by the tremendous amounts of high-impact publications in top-journals dealing with deep learning in medical imaging (for example [16,20,29,31,39,89,136,139,161,272,284], all published in 2018).”
- Recognition in Terra Incognita (2018, methodology): “Automated visual recognition algorithms have recently achieved human expert performance at visual classification tasks in field biology [1–3] and medicine [4, 5].”
- A Survey of Deep Learning for Scientific Discovery (2020, background): “In many supervised learning applications, ranging from machine translation [2] to scientific settings [187, 176], neural networks are trained in a multitask way – predicting several different outputs for a single input.”
- Digital technology, tele-medicine and artificial intelligence in ophthalmology: A global perspective (2020, background): “There needs to be recognition though, that AI may be proven to be more accurate than a physician, and detect features humans cannot, as demonstrated by an algorithm being able to identify sex from fundus photographs (Poplin et al., 2018).”
- Med-BERT: pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction (2020, background): “Given enough training samples, deep-learning models can achieve comparable or even better performance than domain experts in the diagnosis of certain diseases [15-19].”
- A Review of Challenges and Opportunities in Machine Learning for Health. (2018, background): “For example, inferring drug-resistance status in tuberculosis from whole-genome sequencing data 103 , predicting cardiovascular risk from EHR data 64,104 , readmission and mortality risk from longitudinal EHR data 39,105 , and predicting cardiovascular risk from retinal images 106 .”
- A guide to deep learning in healthcare (2019, background): “…of diagnostic tasks, including identifying moles from melanomas 9,10 , diabetic retinopathy, cardiovascular risk, and referrals from fundus 15,16 and optical coherence tomography (OCT) 17 images of the eye, breast lesion detection in mammograms 13 , and spinal analysis with magnetic…”
- Embedding Human Knowledge into Deep Neural Network via Attention Map (2019, methodology): “To achieve an automatic diagnosis, medical image recognition has been attempted for various recognition tasks, such as retinal disease recognition (Jeffrey et al., 2018) and risk forecasting of heart disease (Ryan et al., 2018).”