Download my CV in doc or pdf form, or view it in html.

Research Plan

My research interests include:

Image segmentation and registration, especially clinically-driven applications
Medical image segmentation systems designed to appeal to the clinical user
3D image analysis and visualization
Interactive systems for medical image analysis and visualization
Computer vision
Machine learning
Medical device design

My dissertation research has focused on the development of a novel image analysis framework, called Shells and Spheres (S&S), which facilitates the development of image analysis algorithms that utilize both statistical and shape-based information. This inherently n-dimensional framework makes use of spherical operators of independently variable radius, placed at each voxel in an image. The primary goal of S&S analysis is to determine the proper size of each sphere such that it reaches, but does not cross, the nearest object boundary, thus allowing the radii of all spheres to form a distance map of the image. These operators can combine to model complex contours, and form statistically significant populations of pixels within objects that can be compared to delineate independent image objects.

I have implemented the framework using the Insight Toolkit (ITK), a cross-platform imaging toolkit developed through the National Library of Medicine (NLM) which our lab has been involved with as toolkit developer since its inception. I have designed an algorithm for medical image segmentation using the S&S framework, and have implemented a multi-dimensional, cross-platform software application for testing and experimentation. I am currently working on integrating my software into ITK-SNAP, an expertly designed visualization and interaction system, to create a clinically appealing tool for automated 3D segmentation.

A large part of my dissertation is focused on further development of this system for application to automated segmentation of the RVOT. I designed and implemented software systems for optimizing algorithm parameters for segmentation, developing a novel technique of utilizing a 2D slice selected from a 3D data set to optimize algorithm parameters for an inherently 3D segmentation. This technique allowed parameter optimization to be performed with only a single manual 2D tracing, rather than a training set of manual 3D segmentations. In addition to reducing the burden of time and technical knowledge on the user, this feature allows efficient optimization of algorithm parameters for each individual 3D data set.

The S&S system has been applied to a number of promising collaborative research endeavors. Preliminary testing and validation was performed on Abdominal Aortic Aneurysm (AAA) CT data, with future hopes of modeling the shape variations of AAAs to aid in predicting the likelihood of a rupture. The feature vectors produced by S&S combine medial and boundary information, lending themselves well to mechanical modeling of complex structures. This advantage has lead to a collaboration with Dr. Michael Sacks and Boston Children's Hospital towards the development of a new Tissue Engineered Pulmonary Valve, in which S&S has been utilized to segment the Right Ventricular Outflow Tract (RVOT) and provide feature vectors leading to the mechanical modeling of the RVOT. A collaboration with Robert Tamburo, Ph.D. used S&S to localize amygdala shape differences between elderly control subjects and elderly subjects with depression. Potential projects are being discussed with collaborators that would apply S&S to studies of Right Ventricle (RV) pathologies and further morphological analysis of the amygdala for diagnosis of new disorders.

Aside from S&S and algorithm development, I am interested in applying modern image analysis techniques, with potential modifications and improvements, to clinical problems. I have a particular interest in hybrid segmentation systems, combining shape-based approaches and statistical approaches, but am also interested in registration and other image analysis problems. I hope to use both my clinical and technical training to help bring cutting-edge image analysis technology into the clinic, embedding such systems into the clinical paradigm to improve patient care.

While working towards my Master's Degree in Computer Science, I also developed a version of the Sonic Flashlight™ for use with 3D ultrasound (3DUS). This work was a collaboration with Duke University, where Dr. Stetten was present for the creation of the first 3D ultrasound machine. The 3D Sonic Flashlight design was taken from concept, to computer model (using SolidWorks), to physical prototype for testing with the 3DUS machine developed at Duke.

Design by JSBWebTemplates