Pre-test probability for coronary artery disease in patients with chest pain based on machine learning techniques
Introduction
Obstructive coronary artery disease (CAD) is one of the representative cardiovascular diseases that could lead to sudden death. [[1], [2], [3], [4], [5]] Therefore, a correct and prompt diagnosis of CAD is a crucial component of disease management to reduce the risk of death and improve the quality of life in CAD patients. [2,[5], [6], [7]] Coronary angiography (CAG) and cardiac computed tomography (CT) are the standard tests for the diagnosis of CAD. CAG, an invasive test for CAD diagnosis, is the gold-standard method for CAD diagnosis but has a risk of complications such as access site bleeding. [8] Also, cardiac CT, a non-invasive imaging test, cannot be performed in an emergency. It can also incur high medical costs for patients who do not have CAD. [8] So, the American College of Cardiology (ACC)/American Heart Association (AHA) and the European Society of Cardiology (ESC) guidelines recommend selecting an appropriate pre-diagnosis test for an individual patient according to the CAD probability. [2] Using the pre-diagnosis test to stratify the obstructive CAD risk could be potentially beneficial. Evaluating an individual's pretest probability (PTP) for CAD makes it possible to select patients with chest pain and a high risk of CAD who might benefit the most from a diagnostic procedure such as CAG or cardiac CT. [2,8] Also, it can be used to stratify the risk of CAD among an unselected cohort of patients hospitalized for chest pain. In the present study, we developed and evaluated a model of an individual's PTP for CAD using machine learning (ML) techniques and limited medical information for diagnosing CAD. ML is a collection of technologies that allow artificial intelligence to learn complex rules and identify patterns in multi-dimensional datasets without explicit programming or applying prior assumptions. [7,[9], [10], [11]] Medical big data used for learning can merge many doctors' experiences and ML as the pattern analyzer can efficiently classify high-risk CAD patients requiring percutaneous coronary intervention. [10,11] This study aimed to develop a practical PTP for obstructive CAD in patients with chest pain using ML. The performance of ML-PTP for CAD is compared to the final result of CAG. We used trained physician diagnoses and CAG results obtained from patients with chest pain. Therefore, we assume that these high-performance ML-PTP models for CAD could promptly and correctly classify CAD in a limited diagnostic situation, raising awareness of CAD in patients with chest pain and assisting the decision-making process for doctors.
Section snippets
Study population
The design of the dataset used in this study has been described in previous studies. [12,13] We have used a single-center, prospective, all-comer registry designed to reflect real-world practice since 2004. [12,13] All patients who underwent invasive CAG at Korea University Guro Hospital (KUGH) in Seoul, South Korea, were prospectively enrolled in the KUGH-percutaneous coronary intervention and vasospastic angina registry. A total 10,177 patients with typical or atypical chest pain underwent
Results
The first ML-PTP training and internal validation dataset allocated a total of 8631 subjects [male = 4745 (54.9%), 58.6 ± 12.5 years] who all underwent CAG between 2004 and 2012. The second dataset for external validation for the performance of ML-PTP allocated 1546 subjects [male = 837 (54.1%), 59.6 ± 12.2 years] who underwent CAG between 2013 and 2014 (Fig. 1, Supplement 1).
Obstructive CAD, the primary endpoint of the study, was diagnosed after CAG in a total of 706 (42.9%) subjects in the
Discussion
In this study, we derived a high-performance model for the ML-PTP that can present the CAD risk according to the individual's risk assessment. An ML-PTP with stable performance was derived through internal and external verification, and the performance of ML-PTP was verified using an independent dataset. 1) We derived an ML-based model consisting of three different models according to the subject used to obtain the information, such as chest pain patients (dataset 1), community first medical
Disclosure
The authors have no financial conflicts of interest relevant to the manuscript to disclose.
Sources of funding
This research was supported by a Hanyang University Research Grant.
Acknowledgment
Yung-Kyun Noh and Byoung Geol Choi is partly supported by the National Research Foundation of Korea Grant (NRF/MSIT2017R1E1A1A03070945) and Hanyang University (HY-2019). Seung-Woon Rha is partly supported by Korea University Guro Hospital ‘KOREA RESEARCH-DRIVEN HOSPITALS’ Grant (O1905411).
References (24)
- et al.
Chronic coronary artery disease: diagnosis and management
Mayo Clin. Proc.
(2009) - et al.
2014 ACC/AHA/AATS/PCNA/SCAI/STS focused update of the guideline for the diagnosis and management of patients with stable ischemic heart disease: a report of the American College of Cardiology/American Heart Association task force on practice guidelines, and the American Association for Thoracic Surgery
Preventive Cardiovascular Nurses Association, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons.
(2014) - et al.
Leveraging machine learning techniques to forecast patient prognosis after percutaneous coronary intervention
JACC Cardiovasc. Interv.
(2019) - et al.
Three-year follow-up of patients with acetylcholine-induced coronary artery spasm combined with insignificant coronary stenosis
Int. J. Cardiol.
(2017) - et al.
The impact of myocardial bridge on coronary artery spasm and long-term clinical outcomes in patients without significant atherosclerotic stenosis
Atherosclerosis.
(2018) The use of the area under the ROC curve in the evaluation of machine learning algorithms
Pattern Recogn.
(1997)- et al.
Estimating pre-test probability of coronary artery disease: Battle of the scores in an evolving CAD landscape
JACC Cardiovasc. Imaging
(2019) - et al.
Evolving management paradigm for stable ischemic heart disease patients: JACC review topic of the week
J. Am. Coll. Cardiol.
(2023) - et al.
A comparison of the updated Diamond-Forrester, CAD Consortium, and CONFIRM history-based risk scores for predicting obstructive coronary artery disease in patients with stable chest pain: the SCOT-HEART Coronary CTA Cohort
JACC Cardiovasc. Imaging
(2019) - et al.
The external validity of prediction models for the diagnosis of obstructive coronary artery disease in patients with stable chest pain: insights from the PROMISE trial
JACC Cardiovasc. Imaging
(2018)
Cardiovascular Diseases (CVDs)
European society of cardiology-recommended coronary artery disease consortium pretest probability scores more accurately predict obstructive coronary disease and cardiovascular events than the diamond and forrester score: the partners registry
Circulation.
Cited by (1)
Design and implementation of a smart Internet of Things chest pain center based on deep learning
2023, Mathematical Biosciences and Engineering
- 1
First authors: Byoung Geol Choi and Ji Young Park.