A Non-negative Matrix Factorization for IVUS Image Classification Using Various Kernels of SVM

Abstract

Intravascular ultrasound (IVUS) is a medical methodology and it is a specially constructed catheter with a miniaturized Ultrasound Probe attached to the distal end of the catheter is a medical imaging technique. An efficient method for IVUS image classification using Non-Negative Matrix Factorization (NNMF) and various Support Vector Machine (SVM) kernels are presented in this study. The input IVUS images are given to NNMF for feature extraction and stored in feature database. Finally, SVM kernels like linear, polynomial, quadratic and Radial Basis Function (RBF) are used for prediction and classification of coronary artery lesions and IVUS-based ML algorithms shows good diagnostic performance for identifying ischemia-producing lesions. An IoT based alert is given to the patient’s database cloud that has information of self or blood relation to alert messages in case of emergency artery disease using wearable sensors. The system produces the classification accuracy of 94% by using NNMF and different SVM kernels.KeywordsIVUS image classificationNon-negative matrix factorizationSupport vector machineKernels