Related Experiment Video
Updated: May 12, 2026
07:25
Predicting Amputation using Local Circulating Mononuclear Progenitor Cells in Angioplasty-treated Patients with Critical Limb Ischemia
Published on: September 22, 2020
Predicting complications of percutaneous coronary intervention using a novel support vector method.
Gyemin Lee1, Hitinder S Gurm, Zeeshan Syed
1Department of Electronic and IT Media Engineering, Seoul National University of Science and Technology, Seoul, Republic of Korea. gyemin@seoultech.ac.kr
Summary
A new augmented one-class learning algorithm (OP-SVM) effectively models percutaneous coronary intervention (PCI) complications. This approach improved prediction accuracy compared to traditional methods, showing promise for broader clinical applications.
Area of Science:
- Machine learning in healthcare
- Cardiovascular intervention outcomes
- Predictive modeling in medicine
Background:
- Percutaneous coronary intervention (PCI) can lead to in-laboratory complications.
- Accurate prediction of these complications is crucial for patient safety and procedural success.
- Existing predictive models may not fully capture the complexities of rare event prediction.
Purpose of the Study:
- To evaluate a novel augmented one-class learning algorithm for modeling in-laboratory PCI complications.
- To compare the performance of this new algorithm against traditional classification methods.
- To assess the feasibility and potential clinical utility of the proposed approach.
Main Methods:
- Utilized data from the Blue Cross Blue Shield of Michigan Cardiovascular Consortium (BMC2) registry (2007-2008).
- Trained models using a novel one-plus-class support vector machine (OP-SVM) algorithm to predict 13 PCI complications.
- Compared OP-SVM performance (discrimination, calibration) against logistic regression (LR), one-class SVM (OC-SVM), and two-class SVM (TC-SVM) using 2009 data.
Main Results:
- The OP-SVM algorithm and its cost-sensitive variant demonstrated superior performance.
- Achieved the highest area under the receiver operating characteristic curve for most PCI complications studied.
- Showed significant improvements in Hosmer-Lemeshow chi-squared values and mean cross-entropy error.
Conclusions:
- The augmented one-class learning approach (OP-SVM) enhances discrimination and calibration for PCI complications.
- OP-SVM outperformed logistic regression and traditional support vector machine classifications.
- This novel algorithm holds potential value for predicting adverse events in diverse clinical settings.