AI detects by ECG structural heart disease | The doctor’s daily life | Medical news

An analysis of the electrocardiogram (ECG) by an artificial intelligence tool (AI) would allow the detection. Therefore, routine, of structural cardiopathies hitherto silent. Moreover, This is what the studies of two teams show. Moreover, one of the University of Columbia (New York, United States) and the other of the Imperial College London (London, United Kingdom), respectively published in the journal Nature and in theEuropean Heart Journal. Therefore, The two teams have each developed an AI tool trained over several hundred thousand ECGs. Furthermore, cardiac ultrasounds, correlating very subtle changes in the ECG route to an anomaly or a future risk of heart structural abnormality.

The work of these teams provides tracks for early screening for structural heart disease, such as valvulopathies or cardiomyopathies. Moreover, “We have colonoscopies, mammograms, but we are not equivalent for most heart diseases”, explains Dr. ai detects ecg structural heart Similarly, Pierre Elias, of the team at the University of Columbia.

The IA tool of Americans. Similarly, Echonext, does better than cardiologists – Ai detects ecg structural heart

The team of the University of Columbia, supported by the National Institutes of Health (NIH), has developed an Echonext screening tool, with the objective that ECG data makes it possible to identify patients eligible for cardiac ultrasound. “” Echonext’s deep learning model was drawn over more than 1.2 million ECG-ultrasound pairs from 230,318 patients. The validation study conducted in four hospital systems (10. 862 patients) has shown that the tool was very precise to identify structural heart problems, including heart failure due to cardiomyopathy, valvulopathy, pulmonary hypertension and severe thickening of the heart, from ECG “, detail the authors in a press release from the University of Columbia.

Subsequently, the team carried out an additional validation of their model, “In a more representative cohort of real ai detects ecg structural heart life”. After identifying patients at high risk of structural cardiopathy still undiagnosed. the researchers followed them for a year, “Without their doctors being aware of the use of Echonext”: 55 % of people tried at high risk by AI had been oriented towards an echocardiography by their doctor (s) and, among them, almost 75 % had structural heart disease. Finally. the team compared the performance of 13 cardiologists to those of AI: out of 3,200 ECG, Echonext detected more often a structural anomaly (77 against 64 %) than cardiologists. If the specialists’ scores improved with AI assistance, they were worse than the AI used alone.

The British IA tool predicts even before the structural changes

The London team in collaboration with Shanghai researchers has focused. on the other hand, specifically on unsocied valve insufficiencies. Their AI tool was able to predict. several years in advance, in 69 to 79 ai detects ecg structural heart % of cases, which patients were most likely to develop one from ECG and this, even before the appearance of symptoms or physical changes detectable by ultrasound. The authors explain. in a university press release, that “The anomalies linked to the valves can first manifest itself by very slight modifications of the electrical activity of the heart”.

Their AI model was first drawn from 988,618 ECG and ultrasounds of more than 400,882 patients at Zhongshan Hospital (Shanghai, China). In a separate group of 34. 214 patients from the Beth Israel Deaconess Medical Center (Boston, United States), the tool has shown its effectiveness for early detection of valvulopathies (between 69 and 79 %), among ethnically diverse populations and in various health systems.

“Our work illustrates the advantages of international collaboration in this expanding field, write the authors. By forming the model on an almost exclusively Chinese population. ai detects ecg structural heart then testing it on an American cohort, we can demonstrate that our AI tool can be applied in various countries and contexts around the world. In the end, this means that it has the potential to help even more patients. »» By 2025. the teams hope – with the support of the National Health Service (NHS) – launching other trials that will assess the advantages of the implementation of the IA model in real life in the hospitals of the Imperial College Healthcare NHS Trust and the Chelsea and Westminster Hospital Nhs Foundation.