Efficient Algorithm in Extracting Left Ventricular Ejection Fraction from Isotopic Ventriculography for Heart Pathologies Diagnosis

Abstract

The isotopic ventriculography is one of the most important imaging techniques used for heart pathologies diagnosis and monitoring. It is a qualitative and quantitative exploration offering to clinician the possibility to track the cardiac function. This study proposes a method of segmentation to delimit the left ventricle based on adaptive diffusion flow algorithm. To validate our results, an experimental study is carried out on 50 radionuclide ventriculography sequence of 50 patients (each sequence spans one heart cycle) visually analyzed by an expert. Then, statistical study is carried out in order to evaluate the results by comparing our method to a morphological based segmentation. Results show that the proposed method is better in detecting left ventricular boundaries with a 99.7\% of AUC value and a correlation factor r=0.97 compared to 91.9\% and a correlation factor r=0.37 for the morphological based segmentation.