Researcher | Research Overview
Ultrasound Image Enhancement and Machine Learning:
I have developed a broad set of computational tools for improving the temporal information generated from 3D imaging, specifically creating high frame rate sequences and de-noised volume images of the mitral and aortic valves from 3D echo. The work boosted Ultrasound frame rater from a native 25 volumes a second to over 500 across a single cardiac cycle. This high frame rate ultrasound is very useful for analyzing the dynamics of the heart valve leaflets during their closure, something previously impossible to image. Such high frame rates provide “small motion assumption” necessary to use many computer vision techniques for segmentation and tracking of valvular structures during the heartbeat. These, in turn, provide a novel tool with which to study valve dynamics beyond the traditional Doppler echo views. This work, though the small motion assumption, enables is volume registration. This, in turn, provides a means of stabilizing the images to enhance the clinical view. I have also published work on the application of machine learning (deep neural networks) to the task of detecting and classifying congenital heart diseases in neonates. It is exciting to see how effective these powerful methodologies can be when coupled with our vast collections of clinical data.
My work in extending the dynamic range of ultrasound images using high dynamic range processing algorithms given the best paper award when presented. Using this method ultrasound machines can simultaneously image highly echogenic (bone) and low echogenic (tissues and deep structures) providing a more informative and holistic view to the clinician.
Image Processing and Computer vision:
My PhD thesis work in computer vision and robotics extended energy minimizing active deformable models (a computer vision approach) into the real-time domain for the first time and applied the results directly to robot control. I applied these efficient deformable models to a number of problems from pedestrian and mobile robot tracking to visual servo control of robotic manipulators allowing grasping of previously unknown, by the robot, objects. I applied these methods to the segmentation of 3D real-time ultrasound imagery for minimally invasive intracardiac procedures. In addition to real-time implementation, I have proven theoretical bounds on the accuracy and convergence of my methods with respect to mathematical, geometric definition of smoothness. I showed the efficiency of the second derivative of curvature minimization as a regularizing term for deformable models instead of using curvature directly as was done in prior work.
Surgical and Imaging Tools for Image-Guided Cardiac Interventions:
Image-guided medical procedures are an ideal application domain for surgical robotics research and are a logical extension my work with manipulators, and computer vision guided control. While receiving NIH funding from two consecutive BRP grants I have leveraged my experience to developed clinically useful systems to enabling technology to achieve beating heart repair of congenital heart defects.
User Interfaces:
My first work in user interfaces was quantifying the interface design process for a system used to view confocal microscopy images of rat neurobiology. These large data sets were shared across two sites: one in the US and one in Sweden. One of the problems I advanced was how to build a user interface for simultaneous viewing and manipulation of enormous data sets across large geographic (latency) distances. This work received the best paper and best student paper awards when presented. I have also studied how the sense of touch (haptics) can be electro-mechanically transmitted as part of an interface to surgical instruments and simulators. The transmission of haptic information from the patient to the surgeon through a user interface provides a sense of touch previously unavailable in minimally invasive/robotic procedures. I have demonstrated the use of these haptic systems for remote palpation by integrated an existing haptic display with a manipulator and conducted a user study examining stiffness discrimination ability with tactile and kinesthetic (classic) force feedback components.
Researcher | Research Background
I have over 13 years of experience in translating research advances in computer science into medicine. For many people, computers are just passive boxes that sit on desks. However, for me, the machine is an active part of the world around it. It observes the environment and acts based on its observations and the intentions of users. I develop systems that are not mere tools but are interactive parts of their environment that extend human capabilities. Interfaces must be established between the artificial system and its human collaborator to harness the potential of these systems. Toward that end, my research career has spanned the areas of computer vision, image analysis, robotics, artificial intelligence (specifically machine learning), and user interfaces.
My long-term goal has been to explore the use of computing and robotics to improve clinical applications and ultimately benefit patients. Computer-aided medicine has the potential to change the face of medicine by enhancing clinical skills, expediting procedures, and providing novel diagnostics tools. Unfortunately, over the years I have seen implementation lag far behind the science of computing and its underlying mathematics. For my part, I have tried to address this through collaboration with clinicians to identify areas that could benefit from improved user interfaces, robotics instrumentation, and qualitative information improvements derived from image processing and analysis, machine learning and mathematical modeling.
I use my expertise to support the development of enhancements to cardiac imaging at Boston Children’s Hospital. The long-term goal is that the imaging enchantments and analysis methods I provide will have a positive impact on the diagnosis, preoperative planning, and intervention (through real-time methods) of congenital heart disease. I have found the interdisciplinary nature of computer science and its application to medicine provides a fertile field of potential research and I look forward to continuing exploring the potential that computers offer to advance medicine.