Artificial Intelligence Matching Cardiologist’s Ability to Identify Cardiovascular Disease

A group of researchers has developed an artificial intelligence that can identify signs of myocardial infarction as well as a human cardiologist. With a database of 148 electrocardiograph (ECG) records from patients with myocardial infarction and 52 healthy controls, Nils Strodthoff of the Fraunhofer Heinrich Hertz Institute in Berlin and Claas Strodthoff of the University Medical Center Schleswig-Holstein in Kiel used a sliding-window technique to input data into an artificial neural network in a process that trained the system to detect myocardial infarction.

They used 90% of the data from ECG records to train the system and used the remaining data to test it. In their testing, they found that the artificial intelligence was able to detect myocardial infarction at a similar rate of a cardiologist.

With the current method of detecting myocardial infarction being a physician interpreting an ECG that may be convoluted with noise and skewed data from the emergency room setting, it can often be very difficult to identify the condition. Due to these variable factors, humans have always outperformed machines in previous diagnoses, however the increasing complexity of artificial neural networks has allowed Strodthoff and Strodthoff to create the first artificial intelligence to match human diagnostic capability for myocardial infarction.

One shortcoming of their study is the relatively small population used to train the system, being that most artificial intelligence algorithms require very large data sets for proper function. It will be difficult for the team to amass this large set of data to train the system, however it is a necessary step to ensure that the artificial intelligence is capable of making accurate diagnoses in the wide array of environments the system would be used in.

Source: MIT Technology Review, Image from Medium