Researchers from the University of Nottingham have developed an AI that predicts heart attacks, with an accuracy greater than that of doctors.
Cardiovascular diseases are the first cause of death worldwide, according to the World Health Organization. It is estimated that approximately 20 million people lose their lives due to a heart problem or disease.
Much work remains to be done to diagnose and treat these diseases on time; especially since most cardiovascular diseases can be prevented. For most people, we just have to lead healthy lifestyles (tobacco, diets).
The need for a better system to predict heart attacks
Therefore, for people at high risk of cardiovascular disease, diagnosis and prevention are vital; That is why organizations like the AHA (American Heart Association) have developed guidelines and methods to predict the risk of a patient having a seizure. Using this system, prediction is possible 72.8% of the time.
Not bad, but could be improved. Scientists at the University of Nottingham believe AI may be the key; To do this, they developed four machine learning algorithms, whereby AI reaches conclusions on its own from the data it receives.
Obviously, the greater the amount of data, the better. Nothing less than data from 378 256 patients from across the UK They were used for testing, though not at a stroke.
Hundreds of thousands of records to improve AI that predicts heart attacks
First, the systems used 295,000 data sets to learn and reach conclusions with a predictive model. They then used the rest of the data to test and improve the models depending on the results.
The resulting system became a specialist in predicting heart attacks and other heart problems; came to outperform the AHA system, with an accuracy of between 74.5% and 76.4%. Not only that, but it produced 1.6% fewer false alarms; that is, patients who are diagnosed with cardiovascular disease without having it.
When they tested the system on a registry of 83,000 patients, it was found that this system could have saved the lives of 355 people; patients who died because they were not diagnosed in time. The key is that this system took into account more factors (such as fatal diseases or certain medications) to improve the prediction.
This is a figure to be picked up with tweezers; Obviously, not all diagnoses are the same. But there are reasons for optimism.