Related Experiment Video
Updated: Oct 26, 2025
09:57
Development and Evaluation of 3D-Printed Cardiovascular Phantoms for Interventional Planning and Training
Published on: January 18, 2021
4.2K
Machine learning for detection of stenoses and aneurysms: application in a physiologically realistic virtual patient
G Jones1, J Parr2, P Nithiarasu1
1Faculty of Science and Engineering, Swansea University, Swansea, UK.
Biomechanics and Modeling in Mechanobiology
|August 1, 2021
Summary
Machine learning accurately detects arterial diseases like stenosis and aneurysms using hemodynamic measurements. Tree-based methods show high performance, even with fewer measurements, enabling potential wearable device screening.
Area of Science:
- Biomedical Engineering
- Medical Informatics
- Cardiovascular Research
Background:
- Arterial diseases such as carotid artery stenosis (CAS), subclavian artery stenosis (SAS), peripheral arterial disease (PAD), and abdominal aortic aneurysms (AAA) pose significant health risks.
- Accurate and early detection of these conditions is crucial for effective patient management and improved outcomes.
- Current diagnostic methods may be invasive or not widely accessible for routine screening.
Purpose of the Study:
- To apply and evaluate machine learning (ML) methods for the detection of four major arterial diseases: CAS, SAS, PAD, and AAA.
- To assess the performance of different ML algorithms using a large virtual patient database.
- To determine the impact of reducing the number of hemodynamic measurements on detection accuracy.
Main Methods:
- Utilized a physiologically realistic virtual patient database (VPD) of 28,868 subjects, augmented to include various arterial diseases.
- Trained and tested tree-based ML methods, including Random Forest and Gradient Boosting, for disease detection.
- Quantified performance using the [Formula: see text] score, sensitivity, and specificity based on hemodynamic measurements (pressure and flow-rate).
Main Results:
- Random Forest and Gradient Boosting demonstrated superior performance compared to other ML approaches.
- High [Formula: see text] scores (up to 0.98) and accuracies (>90% sensitivity/specificity) were achieved for CAS, PAD, SAS, and AAA with six hemodynamic measurements.
- Performance degradation was minimal (<10%) even when reducing measurements to two, with single measurements effective for AAA detection.
Conclusions:
- Machine learning, particularly tree-based methods, shows high efficacy in detecting major arterial diseases using hemodynamic data.
- The feasibility of using a limited number of measurements, even a single one for AAA, suggests potential for non-invasive screening.
- These findings support the development of wearable devices for continuous monitoring and early detection of arterial conditions like AAA.
