Artificial intelligence model improves accuracy of arrhythmia diagnosis

Accurate diagnosis and treatment are necessary in cases ventricular tachycardia. However, this care can now be optimized thanks to new artificial intelligence and machine learning systemcapable of pinpointing the location of heart rhythm disturbances in each patient, known scientifically as S.O.O. (place of origin in English).

The project is led by a group of researchers from University of Pompeu Fabra (UPF) and through this they intend increase the effectiveness of techniques used in the treatment of arrhythmias. In addition to prescribed medications aimed at correcting heart rhythm disturbances, radiofrequency ablation This is a method that is currently used primarily to treat heart failure. However, its efficiency still needs to be optimized.

The success of radiofrequency ablation depends on the accuracy of localization of the arrhythmia.

The radiofrequency ablation procedure consists of introduction of radiofrequency catheters counteract changes in heart rhythm so that thermal energy destroys exactly that part of the heart tissue in which the arrhythmia occurs. However, to use ablation, it is necessary to first map the electrical circuit to get an idea of ​​the exact area where the catheter should be placed.

The effectiveness of this treatment method can still be increased, since determining the location of arrhythmia is one of the most important tasks of our time. Therefore, the use of a new artificial intelligence model developed by UPF will help increase efficiency in the moment determine the location of arrhythmia and therefore in the success of radiofrequency ablation. At the same time, this will allow you to achieve reduce intervention time and the likelihood of new arrhythmias.

ALGORITHM DEVELOPMENT

The new system used combining real clinical data using artificial intelligence and machine learning (machine learning) such as the age, gender and medical history of the patient, paying special attention to whether the patient is hypertensive or not. And also from the analysis of real and simulated electrocardiograms using computational methods. This will make it easier to diagnose where the heart rhythm disorder has occurred and improve the effectiveness of treatment.

arrhythmias ventricular They begin in the inner chambers of the heart, the ventricles, but come in different types. The UPF project focuses on ventricular outflow tract arrhythmias (VAT), that is, the area connecting the ventricles with the main arteries of the heart. This type of arrhythmia is the most common of the idiopathic ventricular arrhythmias, and it is its causes that cannot be identified by traditional methods or in patients who do not have structural heart disease.

Artificial intelligence (AI) and machine learning expand the capabilities of the human eye

Traditional methods for detecting arrhythmias are fundamentally based on analysis electrocardiogram (ECG), which are performed on the patient before surgery to treat arrhythmia. These electrocardiograms are visually interpreted by medical personnel and can be subject to human error. Consequently, there is a potential for erroneous or inaccurate diagnoses that impact patient care and treatment effectiveness.

Although several models based on artificial intelligence and machine learning have already been launched in recent years to expand the limitations that the human eye presents in the analysis of electrocardiograms, the truth is that they have not been able to accurately determine the site of origin or SOO. arrhythmias. To overcome this obstacle, the UPF system used unsupervised hierarchical clustering method explore the internal organization of data.

The results obtained using the artificial intelligence and machine learning model achieved a good level of sensitivity and accuracy.

Unlike other approaches aimed at distinguishing the sources of right ventricular (RVOT) and left ventricular (LVOT) arrhythmias, the developed artificial intelligence model was designed to discover specific place of origin (SOO) ventricular arrhythmia in the structures of the heart. This innovative methodology deepens the understanding of ventricular arrhythmias and refines the direction of treatment strategies by incorporating into the algorithm various databases, supervised and unsupervised learning models, feature and pattern analysis, and various experiments performed.

The results obtained reached good level of sensitivity and accuracy. After the experiments, the lowest values ​​found were 61% accuracy and 41% sensitivity, although some tests showed higher values. For now, the research team hopes feed the algorithm with the help of additional analysis, expand the collected characteristics and collect more cases in order to develop a more reliable and efficient system in identifying the location of heart rhythm disturbances.