Learning sparse kernels from 3D surfaces for heart wall motion abnormality detection

摘要: Coronary heart disease (CHD) is a global epidemic that the leading cause of death worldwide. CHD can be detected by measuring and scoring regional motion left ventricle (LV) heart. This project describes novel automatic technique which detect wall abnormalitie, LV from echocardiograms. Given sequence endocardial contours extracted ultrasound images, moving through time interpreted as three-dimensional (3D) surface. From 3D surfaces, we compute several geometry-based features (shape-index values, curvedness, surface normals, etc.) to obtain histograms-based similarity functions are optimally combined using mathematical programming approach learn kernel function designed classify normal vs. abnormal motion. In contrast with other state-of-the-art methods, our formulation also generates sparse kernels. Kernel sparsity directly related computational cost evaluation, an important factor when designing classifiers part real-time system. Experimental results on set echocardiograms collected in routine clinical practice at one hospital demonstrate potential proposed approach.