Spatio-temporal analysis of three-dimensional real-time ultrasound for quantification of ventricular function.

摘要

Among screening modalities, echocardiography is the fastest, least expensive, and least invasive method for imaging the heart. A new generation of three-dimensional ultrasound technology has been developed with real-time three-dimensional (RT3D) matrix phased array transducers. These transducers allow interactive three-dimensional visualization of cardiac anatomy and fast ventricular volume estimation without tomographic interpolation as required with earlier 3D ultrasound acquisition systems. However, real-time acquisition speed is performed at the cost of decreasing spatial resolution leading to echocardiographic data with poor definition of anatomical structures and high levels of speckle noise. The poor quality of the ultrasound signal has limited the acceptance of RT3D ultrasound technology in clinical practice, despite the wealth of information acquired by this system, far greater than with any other existing echocardiography screening modality. This dissertation aimed to improve the better acceptance of this new technology by addressing the problem of automatic quantification of ventricular function.; Denoising of RT3D ultrasound was applied as a preprocessing step to improve image quality using spatio-temporal brushlet basis functions to characterize echocardiographic data in terms of oriented texture components and decorrelate non-coherent speckle noise in the frequency domain. Denoising experiments on phantom and clinical data showed that brushlet analysis was well adapted to the intrinsic nature of RT3D ultrasound data and performed better than traditional denoising methods. Experiments also showed that including the time dimension directly in a brushlet expansion exploited temporal coherence between successive frames to identify cardiac structures while removing speckle noise components, not correlated in time. This dissertation also identified the set of ‘best’ parameters to optimize denoising performance both visually and quantitatively with signal-to-noise ratio measurements.; Deformable-model segmentation methods were implemented in two dimensions using a parametric formulation and in three dimensions using an implicit formulation with a level set implementation for extraction of endocardial surfaces on denoised RT3D ultrasound data. A complete and rigorous validation of the segmentation methods was carried out for quantification of left and right ventricular volumes and ejection fraction including comparison of measurements with cardiac MRI as the reference.