Artificial intelligence equals cardiologists in the detection of heart attacks

Nils Strodthoff, researcher at the Fraunhofer Institute Heinrich Hertz (Germany) and Claas Strodthoff, at the Medical Center of the University of Schleswig-Holstein (Germany) developed a neural network capable of detecting the signs myocardial infarction. is equal to the performance of human cardiologists.

A standard electrocardiograph records the electrical signal through 12 different wires connected to different parts of the patient's body. These signals denote the electrical behavior of the heart in different ways. However, the interpretation of the data is difficult. The clbadification process is further complicated by the presence of ST segment elevation (within an electrocardiogram). Patients who have such elevation should be treated as soon as possible, while those who do not have it need more tests that take a lot of time.

Strodthoff and Strodthoff fed a neural network with a database of 148 electrocardiographic (ECG) records of patients with myocardial infarction and 52 healthy controls. They used a sliding window technique to transmit this data to a neural network. Each window had at least three pulsations. The team used 90% of the data to form this network and learned to detect the signs of a heart attack.

The rest of the data was used to test if the network was working and they got some interesting results. "The proposed architecture overcomes the latest state-of-the-art approaches in this dataset and reaches a level of performance similar to that of human cardiologists," say Strodthoff and Strodthoff.

In addition, machines tend to rely on same signal data as cardiologists. This impressive result shows the potential of smart machines to radically improve medical care.

Machine learning algorithms require large sets of data to learn correctly, but this is not the case when many physicians are suspicious of the results. Generating larger sets of data from cardiac arrest registrations will be slow and difficult, but it is only then that doctors will be sure that the vast array of drugs will be available. the chaotic environments in which they work is covered.

The approach developed by Strodthoff and Strodthoff is generically applicable to any clbadification problem in time series of raw data from devices such as the electrocardiogram and electroencephalography (EEG), which abound in medicine, which opens up a vast panorama of their uses in health.