Cardiac resynchronization therapy (CRT) is the currently used non-pharmacologic therapy for heart failure (HF) patients with a conduction disturbance. The purpose of CRT is to resynchronize contraction between and within ventricles. This dyssynchrony arises from an abnormality in the heart's electrical conducting system that causes the two ventricles to beat asynchronously. Studies from clinical trials have shown that about 30% of the patients do not gain any benefits from CRT. Computational models may have the potential to predict the response of patients to CRT and could be used to optimize the effects of this therapy. Hence, we have developed methods for generating anatomically detailed patient-specific three-dimensional finite element models of the failing human heart from clinical measurements. Two three-dimensional finite element models were generated from cardiac computed tomography and magnetic resonance images obtained at end-diastole using a newly developed semi-automatic technique of landmark points. Landmark points are anatomical locations identified on the cardiac images that allow us to estimate the extent of the heart in order to build a 3D model of the failing heart. Seventeen landmark points were used to precisely determine the circumferential and longitudinal extent of the failing human heart. A 3D finite element model with 50 nodes and 24 elements was generated for two patients with dyssynchronous heart failure. These 3D models were then used to estimate myocardial wall thickness and ventricular cavity volume. The method of landmark points provides an efficient way to generate finite-element meshes.