Automated Classification of Disease Patterns from Echo-cardiography Images Based on Shape Features of the Left Ventricle
- School of Education Technology, Jadavpur University, Kolkata 700032 (India)
Computer assisted diagnosis using analysis of medical images is an area of active research in health informatics. This paper proposes a technique for indication of heart diseases by using information related to shapes of the left ventricle (LV). LV boundaries are tracked from echo-cardiography images taken from LV short axis view, corresponding to two disease conditions viz. dilated cardiomyopathy and hypertrophic cardiomyopathy, and discriminated from the normal condition. The LV shapes are modeled using shape histograms generated by plotting the frequency of normalized radii lengths drawn from the centroid to the periphery, against a specific number of bins. A 3-layer neural network activated by a log-sigmoid function is used to classify the shape histograms into one of the three classes. Experimentations on a dataset of 240 images show recognition accuracies of the order of 80%.
- OSTI ID:
- 21428704
- Journal Information:
- AIP Conference Proceedings, Vol. 1298, Issue 1; Conference: ICMOS 20110: International conference on modeling, optimization and computing, West Bengal (India), 28-30 Oct 2010; Other Information: DOI: 10.1063/1.3516368; (c) 2010 American Institute of Physics; ISSN 0094-243X
- Country of Publication:
- United States
- Language:
- English
Similar Records
First third filling parameters of left ventricle assessed from gated equilibrium studies in patients with various heart diseases
Transient postexercise ischemic dilatation of the left ventricle on stress-redistribution T1-201 scintigrams: Marker of severe and extensive coronary artery disease