New AI system could revolutionize diagnosis of valvular heart disease

AI (artificial intelligence) may sound like a cold robotic system, but scientists at Osaka Metropolitan University have shown that AI can provide heart-warming, or even “mind-warning” support. rice field. They announced a groundbreaking use of AI to classify cardiac function and pinpoint valvular heart disease with unprecedented precision, and continue to merge the fields of medicine and technology to advance patient care. demonstrated significant progress. Result is, lancet digital health.

Valvular heart disease, one of the causes of heart failure, is often diagnosed using echocardiography. However, there is a shortage of qualified technicians as this technology requires specialized skills. A chest x-ray, on the other hand, is one of the most common tests, primarily for identifying lung disease. The heart can also be seen on chest radiographs, but until now little was known about the ability of chest radiographs to detect heart function and disease. Chest radiography (chest x-ray) is performed in many hospitals, takes little time to perform, is accessible and reproducible. Therefore, a research team led by Professor Taiki Ueda of the Department of Diagnostic Radiology, Graduate School of Medicine, Osaka Metropolitan University, thought that this test would be useful if it were possible to determine heart function and disease from chest X-rays. Supplement to echocardiography.

Dr. Ueda’s team has successfully developed a model that uses AI to accurately classify heart function and valvular heart disease from chest X-rays. The team aimed at data from multiple institutions, as an AI trained on a single dataset would face potential biases, leading to poor accuracy. Thus, between 2013 and 2021, a total of 22,551 chest radiographs associated with 22,551 echocardiograms were collected from 16,946 patients at four institutions. With chest radiographs set as input data and echocardiograms set as output data, an AI model was trained to learn relevant features. both datasets.

The AI ​​model was able to accurately classify the six selected types of valvular heart disease with area under the curve (AUC) ranging from 0.83 to 0.92. (AUC is a metric that indicates the ability of the AI ​​model and uses a range of values ​​from 0 to 1, with values ​​closer to 1 being better.) AUC was 0.92. Fraction – an important measure for monitoring heart function.

“It took a very long time to reach this result, but I believe it is a meaningful study,” said Dr. Ueda. He said, “In addition to improving the diagnostic efficiency of doctors, there is a possibility that it can be used in areas where there are no specialists, in emergency situations at night, and for patients who have difficulty in echocardiography.”