Researchers have created an AI algorithm capable of precisely diagnosing heart functions and valvular heart diseases through the use of chest X-rays. This advancement may serve as a complement to conventional echocardiography and is particularly beneficial in environments that lack specialized personnel.
Scientists introduce innovative and precise AI-enabled techniques for evaluating cardiac functionality and conditions via chest radiography.
While artificial intelligence (AI) is often thought of as a devoid-of-emotion, machine-oriented system, scholars at Osaka Metropolitan University have demonstrated its capacity to offer not just heartwarming, but crucially, “heart-alerting” aid.
The research group has engineered a pioneering AI application that effectively categorizes cardiac functions and diagnoses valvular heart diseases, marking continued progress in the merging of medical science and technological innovation to enhance patient care. The results have recently been published in the scholarly journal, The Lancet Digital Health.
The diagnosis of valvular heart disease, a contributor to heart failure, is commonly performed via echocardiography. This method demands specialized expertise, resulting in a corresponding deficit of skilled technicians. In contrast, chest X-rays are a prevalent diagnostic tool primarily utilized for detecting lung conditions. Despite the visibility of the heart in these images, their potential for assessing cardiac functionality or ailments has been largely unexplored until now.
Left: Chest X-ray. Right: Graphical representation of the basis for the AI’s analysis. Credit: Daiju Ueda, OMU.
Chest radiographs are conducted in numerous healthcare facilities and require minimal time, making them a highly convenient and replicable test. Thus, the research team, spearheaded by Dr. Daiju Ueda from the Department of Diagnostic and Interventional Radiology at the Graduate School of Medicine of Osaka Metropolitan University, posited that chest X-rays could augment echocardiography in determining cardiac conditions.
The team successfully developed a model employing AI to unerringly categorize cardiac functionalities and valvular heart diseases based on chest X-rays. Recognizing the potential for bias and lower accuracy in an AI model trained on a singular dataset, the researchers targeted multi-institutional data. To this end, they amassed a total of 22,551 chest X-rays paired with 22,551 echocardiograms from 16,946 patients across four institutions between 2013 and 2021. These X-rays served as the input, and the echocardiograms as the output, for training the AI algorithm to identify relevant features across both datasets.
The AI model demonstrated a high level of precision in diagnosing six specific types of valvular heart diseases, with an Area Under the Curve (AUC) that ranged from 0.83 to 0.92. (The AUC is a performance metric that ranges from 0 to 1; the closer to 1, the better the model’s capability.) Notably, the AUC reached 0.92 at a 40% cut-off for identifying left ventricular ejection fraction, a critical parameter for monitoring cardiac health.
“It took an extensive period to arrive at these findings, yet they are profoundly impactful,” remarked Dr. Ueda. “Beyond augmenting the diagnostic efficacy of medical professionals, this technology may find applications in areas devoid of specialists, during nocturnal emergencies, or for patients for whom echocardiography is challenging.”
Reference: “Artificial Intelligence-Based Model to Classify Cardiac Functions from Chest Radiographs: A Multi-Institutional, Retrospective Model Development and Validation Study” by Daiju Ueda, Toshimasa Matsumoto, Shoichi Ehara, Akira Yamamoto, Shannon L Walston, Asahiro Ito, Taro Shimono, Masatsugu Shiba, Tohru Takeshita, Daiju Fukuda and Yukio Miki, published on July 6, 2023, in The Lancet Digital Health.
DOI: 10.1016/S2589-7500(23)00107-3
Table of Contents
Frequently Asked Questions (FAQs) about AI-based cardiac diagnosis
What is the primary aim of the research conducted by Osaka Metropolitan University?
The primary aim is to develop an AI model capable of accurately diagnosing heart functions and valvular heart diseases using chest X-rays. This could complement existing diagnostic methods like echocardiography, particularly in settings that lack specialized technicians.
Who led the research team at Osaka Metropolitan University?
Dr. Daiju Ueda from the Department of Diagnostic and Interventional Radiology at the Graduate School of Medicine led the research team at Osaka Metropolitan University.
Where were the findings of this research published?
The findings were published in the scholarly journal, The Lancet Digital Health.
What makes this AI model groundbreaking?
The AI model is groundbreaking because it can accurately categorize cardiac functionalities and diagnose specific types of valvular heart diseases using chest X-rays. This opens the possibility of using a commonly available diagnostic tool to augment more specialized techniques like echocardiography.
How did researchers ensure the accuracy of the AI model?
To ensure the accuracy, the research team used multi-institutional data comprising 22,551 chest X-rays and corresponding echocardiograms from 16,946 patients across four institutions, collected between 2013 and 2021.
What is the Area Under the Curve (AUC) in this context?
The Area Under the Curve (AUC) is a performance metric that indicates the capability of the AI model to make accurate diagnoses. In this study, the AUC ranged from 0.83 to 0.92, indicating high diagnostic accuracy.
How can this technology be utilized in real-world medical settings?
Apart from improving diagnostic efficiency among medical professionals, this technology can also be useful in areas that lack specialized technicians, for emergencies during night-time, and for patients who may have difficulty undergoing traditional echocardiography.
What types of cardiac conditions can the AI model diagnose?
The AI model can accurately diagnose six specific types of valvular heart diseases and can also categorize cardiac functionalities like left ventricular ejection fraction.
How long did it take for the research team to develop this AI model?
The research took an extensive period to complete, although the exact duration is not specified in the text. Dr. Daiju Ueda stated that the results are profoundly impactful, indicating that the time invested was substantial.
More about AI-based cardiac diagnosis
- The Lancet Digital Health Journal
- Osaka Metropolitan University Official Website
- Introduction to Echocardiography
- Area Under the Curve (AUC) Explained
- Valvular Heart Disease Overview
