Spectrum-Based Automatic Localization of Left Ventricle from Cardiac Magnetic Resonance Imaging

Liang Zhong¹, Min Wan¹, Jun Mei Zhang¹, Ru San Tan¹

A novel method for automatic localization of left ventricle (LV) via spectrum analysis in magnetic resonance imaging is proposed. Two temporal approaches based on the variance image and the first harmonic image are applied and reformulated in a unified framework to reveal their intrinsic connections. The first harmonic image and some higher harmonic image are used in an anisotropic weighted circle Hough transform. The first harmonic image votes with positive weights while the higher harmonic image votes with negative weights to suppress the right ventricular interruption. The results in 17 patients show the capability for automatic localization of LV.