Open AccessCardiac Motions Classification on Sequential PSAX Echocardiogram
Author(s) -
Adam Shidqul Aziz,
Riyanto Sigit,
Achmad Basuki,
Taufik Hidayat
Publication year - 2018
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v12.i3.pp1289-1296
Subject(s) - artificial intelligence , correctness , computer science , endocardium , displacement (psychology) , noise (video) , segmentation , pattern recognition (psychology) , boundary (topology) , speckle noise , computer vision , mathematics , speckle pattern , algorithm , image (mathematics) , medicine , cardiology , psychology , mathematical analysis , psychotherapist
Cardiac wall motions classification on 2-dimensional (2D) echocardiographic images is an important issue for quantitative diagnosiing of heart disease. Unfortunately, the bad quality of echocardiogram cause computationally classification on cardiac wall motions is still become a big homework for many researchers to provide the best result. Echocardiogram is produced by soundwaves which absolutely make its images have speckle noise in different intensity. Therefore, this research improves a set of methodology to classify cardiac wall motion semi-automatically. Raw echocardiogram will be enhanced and segmented to take the boundary of endocardium of left ventricular in PSAX cardiac images. New improvement of Semi-automatically methodology is approach on detecting the contour of endocardium and will be inputed as good features in Lucas-Kanade Optical Flow in all sequential echocargraphic images. On classifying cardiac wall motions, this research proposes two important features including length of displacement and flow direction. New proprosed flow determination algorithm and Euclidean distance is used to calculate those features. All the features will be trained by Neural Network (NN) and validated by Leave One Out (LOO) to get accurate result. NN method, which is validated by LOO, has the best result of 81.82% correctness than the other compared methods.