How does artificial intelligence work? Mainly it uses algorithms to predict patterns. A new algorithm that predicts heart attacks, called LogitBoost, is more accurate than current practices. It was announced on May 12, 2019, at the International Conference on Nuclear Imaging and Cardiac CT in Lisbon, Portugal.
How Cardiac PET Imaging Helped
A study was done over six years comparing real patients and their outcomes to the new algorithm. A human doctor takes all of the variables of a patient and decides if that patient is at risk for a lethal heart attack. Variables are factors to consider, such as patient’s age, gender, if he or she smokes or has diabetes, and so on.
Some of the variables included results from scans of the heart, including cardiac PET scans. The study noted that a coronary computed tomography angiography (CCTA) scan helped show 58 variables on the circulatory system, such as if any blood vessels were narrowing. A PET scan produced seventeen variables related to how well a patient’s blood was flowing through his or her body.
950 patients with chest pain were entered into a study. They received CCTA scans and those showing risks for coronary artery disease were given further cardiac PET imaging scans. After six years, 24 had heart attacks, and 85 were dead. Could LogitBoost predict these results better than the human doctors could?
The answer is a resounding yes. LogitBoost made a 90 percent accurate prediction, far better than the doctors of the study patients. It is thought that LogitBoost can compare results between PET and CCTA scans that a person can miss because he or she is only human.
So What Does This Study Mean?
Although further studies are needed to test LogitBoost’s accuracy, it is hoped that this can help doctors get better help more quickly to patients complaining of chest pains.