Recovery of Heart Material and Kinematics Properties from Image Sequences

Cardiovascular diseases remain the leading cause of death in the world. It has been recognized that myocardial ischemia often manifests as the morphological and kinematic abnormalities of the left ventricle, as well as alterations in myocardial fiber structures and material elasticity. The intent of the effort is to develop an image-based computational framework, which uses various computer vision and system estimation techniques along with stochastic inverse mechanics principles, to study the material and kinematics properties of deformable objects, with particular applications to the diagnosis of heart conditions. In addition to the non-invasive diagnostic utility of the results, this framework contributes to the basic understanding of cardiac physiology by bridging the complementary advantages of medical imaging and biomechanics approaches. It also contributes to the general areas of model-based computer vision, and model-based system and parameter estimation.

 

Spatiotemporal Evolution Models for Joint Shape and Motion Analysis

In the quantitative analysis of cardiac image sequence, tissue boundary extraction and motion recovery are two of the essential objectives, which are traditionally treated as two sequential steps. However, it is advantageous to treat the spatial boundary finding and temporal correspondence tracking as a coherent and unified process such that the image information is more fully used, and the result is more robust. We aim to develop a unified spatiotemporal evolution paradigm that can simultaneously recover the boundary shape and the motion information of the heart from image sequence.

 

Computer-Aided Detection and Diagnosis of Mammogram

We are exploring the possibilities of developing comprehensive computer-aided screening and detection strategies that extract and analyze the characteristics of benign and malignant lesions from mammographic images. The emphasis is on the detection of micro-calcifications, the earliest signs of breast cancers.

 

Virtual Reality Based Systems For Training On Endoscopic Surgery And Diagnostic Obstetric Ultrasound Procedures

The use of new medical technology requires the acquisition of new skill, which could be difficult or undesirable to be obtained in real clinical settings. Two important examples of these techniques are endoscopic surgery and obstetric ultrasound examination. Virtual reality based simulation systems provide a very elegant solution to the problem, because we can provide virtual models of different anatomic structures to simulate different procedures in realism within the virtual environment.

.

 

Directional Wavelet Analysis of Edges

Numerous edge detection algorithms have been proposed in the past thirty years with varying degrees of success. We are working on a signal processing approach to the problem by analyzing and extracting edges from 2D images using directional wavelets analysis. Edge detection can be greatly improved by identifying the type of edge that is being analyzed and selecting/constructing the most appropriate wavelet basis and scale for the analysis. In addition, wavelet bases with directional information, such as complex wavelets, offer possible advantages because they can conform to edge directions.