ANO Detection with K-Nearest Neighbor Using Minkowski Distance
Hui-Lee Ooi, Siew-Cheok Ng, and Einly Lim
University of Malaya/ Department of Biomedical Engineering, Kuala Lumpur, Malaysia
Abstract—Full unloading of the left ventricle in patients implanted with ventricular assist devices over a long period of time has been reported to cause aortic valve fusion and thrombosis. Therefore, it is vital to detect the occurrence of aortic valve non-opening (ANO) in these patients. In the present study, measurements were obtained from four greyhounds under a wide range of operating conditions, which covered varying levels of blood volume, vascular resistances, contractilities and pumping speeds. K nearest neighbor classifier was implemented on two features, i.e. root mean square and standard deviation of pump speed amplitude. Classification performance using different Minkowski distance metrics (Manhattan distance, Euclidean distance, Minkowski distance with a power of 3 and Chebyshev distance) and k parameter (odd numbers ranging from 1 to 29) were evaluated. Results showed that the compared metrics achieved similar performance (accuracy of 93%) and concurred unanimously with regards to the optimal numbers of k parameters (k=9) for the data tested.
Index Terms—aortic valve non-opening, ANO detection, k-nearest neighbor, Minkowski distance
Cite: Hui-Lee Ooi, Siew-Cheok Ng, and Einly Lim, "ANO Detection with K-Nearest Neighbor Using Minkowski Distance," International Journal of Signal Processing Systems, Vol. 1, No. 2, pp. 208-211, December 2013. doi: 10.12720/ijsps.1.2.208-211