Heart conditions, such as heart attacks, are among the top causes of death globally. Early diagnosis of the likelihood of a heart attack is important in order to provide early medical treatment, which can greatly enhance the outcome of care. Recent advances in artificial intelligence and imaging technology have created new hopes of evaluating the health of the heart without having to go through invasive medical procedures. This study examines whether risk of heart attack can be forecast from retinal images—images of the retina of the eye. Since the retina reflects the state of one's vessels in the entire body, the state of one's cardiovascular system is possible to ascertain from such inspection. To do so, we created a deep learning model that could detect patterns in retinal images associated with heart disease risk. We trained and tested the model using a big database of retinal images and correlated it with heart health information. We discovered that this automated method can detect risk of heart attack with accuracy in a completely non-invasive and inexpensive way. Using retinal imaging and machine learning, this method is capable of transforming the early detection of heart disease such that it helps patients make provisions and eventually end the prevalence of heart disease around the world.