A machine learning model using echocardiographic myocardial strain to detect myocardial ischemia

Coronary functional assessment plays a critical role in guiding decisions regarding coronary revascularization. Traditional methods for evaluating functional myocardial ischemia, such as invasive procedures or those involving radiation, have their limitations. Echocardiographic myocardial strain has...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Internal and emergency medicine Ročník 20; číslo 5; s. 1425 - 1436
Hlavní autori: Zheng, Bo, Liu, Yaokun, Zhang, Jingyi, Ma, Terry T., Zhou, Yun, Chen, Yongkai, Yang, Ying, Ma, Wei, Fan, Fangfang, Jia, Jia, Zhang, Yan, Li, Jianping, Zhong, Wenxuan
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Cham Springer International Publishing 01.08.2025
Springer Nature B.V
Predmet:
Aged
Angiography
Annotations
Approximation
Artificial intelligence
Automation
Cardiovascular disease
Coronary artery disease
Coronary vessels
Echocardiography - methods
Ejection fraction
Female
Heart
Heart diseases
Humans
Hypertension
Im - Original
Internal Medicine
Ischemia
Learning algorithms
Machine Learning
Male
Medical imaging
Medicine
Medicine & Public Health
Middle Aged
Myocardial ischemia
Myocardial Ischemia - diagnosis
Myocardial Ischemia - diagnostic imaging
Myocardial Ischemia - physiopathology
Pilot Projects
Pulmonary arteries
Quality standards
ROC Curve
Ultrasonic imaging
Myocardial ischemia
Ensemble learning
Echocardiographic myocardial strain
Coronary angiography-derived fractional flow reserve
Machine learning
ISSN:1828-0447, 1970-9366, 1970-9366
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Coronary functional assessment plays a critical role in guiding decisions regarding coronary revascularization. Traditional methods for evaluating functional myocardial ischemia, such as invasive procedures or those involving radiation, have their limitations. Echocardiographic myocardial strain has emerged as a non-invasive and convenient indicator. However, the interpretation of strain values can be subject to inter-operator variability. Artificial intelligence (AI) and machine learning techniques may promise to reduce the variability. By training AI algorithms on a diverse range of echocardiographic data, including strain values, and correlating them with ischemia, it may be possible to develop a robust and automated diagnostic tool. This study aims to provide a non-invasive and effective solution for automated myocardial ischemia detection that can be used in clinical practice. To construct the machine learning model, we used an automatic left ventricular endocardium tracing tool to extract myocardial strain data and integrated it with six clinical features. A coronary angiography-derived fractional flow reserve (caFFR) ≤ 0.80 was defined as the indicator of myocardial ischemia. A total of 636 suspected coronary artery disease subjects were enrolled in this pilot study, where 282 cases (44.3%) had myocardial ischemia. These subjects were randomly divided into training ( n  = 508) and testing ( n  = 128) sets at a 4:1. Using ensemble-learning algorithms to train and optimize the model, its diagnostic performance versus caFFR was diagnostic accuracy 85.9%, sensitivity 88.9%, specificity 83.1%, positive predictive value 83.6%, negative predictive value 88.5%. The optimized model achieved an area under the receiver operating characteristic curve (AUC) of 0.915 (95% confidence interval [CI] 0.862–0.968). Our machine learning prototype model based on echocardiographic myocardial strain shows promising results in detecting myocardial ischemia. Further studies are needed to validate its robustness and generalizability on larger patient populations. Graphical Abstract Summary of the machine learning model for detecting myocardial ischemia based on echocardiographic myocardial strain. CAD coronary artery disease, caFFR coronary angiography-derived fractional flow reserve, AUC area under the receiver operating characteristic curve
Bibliografia:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:1828-0447
1970-9366
1970-9366
DOI:10.1007/s11739-025-03968-6