I to tackle other major diseases, including autism and

I plan to dedicate my career to advancing computational medicine, which I believe can
dramatically improve the quality and efficiency of our health care. I am fascinated with machine
learning (ML) in the context of medical image analysis, and as an ML researcher in the Yale
Radiology Research Lab working under Professor James Duncan, I have helped develop
algorithms that could improve the quality and speed of diagnostic and interventional radiology
readings. I am pursuing a PhD in Biomedical Engineering in order to gain the opportunities and
robust training necessary to achieve my aspirations in the field.
Working with Dr. Duncan has allowed me to pursue challenging and clinically relevant problems
in liver cancer imaging, including automatic classification and segmentation of lesions from
abdominal MRIs using ML. Yale’s PhD program would help me contribute further to these
research areas with the confidence that I will continue to benefit from the department’s
productive environment, multidisciplinary collaborations and strong research opportunities. I am
also excited by the rapid progress Dr. Duncan’s lab has made in incorporating and improving
novel approaches such as neural networks in its efforts to tackle other major diseases, including
autism and coronary artery disease.
As part of my graduate studies, I would be interested in identifying multi-timepoint MRI-derived
features related to lesion shape, appearance and microenvironment that predict response of liver
cancer to locoregional therapy in pre-operative imaging or predict cancer recurrence in postoperative
imaging. As well, I would like to build on my current work in deep learning to
investigate transfer learning in the context of medical imaging; for example, neural network
representations of liver lesions for use in diagnosis could be reused for applications in staging,
prognosis, and response prediction and assessment. A successful demonstration of transfer
learning could greatly accelerate progress in medical ML by reducing the effort, data and
computational power required to design and train models for different diseases and image
analysis tasks.
My programming and research experiences give me the confidence that I will be a valuable asset
to Dr. Duncan’s group. By the time I start my PhD program, I will have a year of full-time
research experience in his lab using ML techniques to address similar medical imaging
challenges, in addition to my undergraduate research experience in biomedical computation as
well as analytics experience in industry. With each new research experience, I have learned a
new set of mathematical and/or coding skills alongside a new biomedical context. Each time, I
have risen to the challenge of independently learning and applying these concepts to establish
meaningful insights.
My ability to learn quickly and independently has been one of my greatest strengths throughout
my academic career, and helped me significantly in my research on describing the binding
dynamics and function of the cardiac muscle protein MyBPC. I needed to develop a model that
captures nonlinear binding interactions of proteins without well-specified kinetic rates. It also
needed to account for varying spatial distributions as well as dynamics along multiple time
scales. This proved to be a large conceptual and computational challenge, requiring me to
integrate knowledge and methods from multiple fields, create and execute a roadmap for tacklingĀ a multifaceted problem, and build and manage a complex code base across multiple languages.
As I had not previously had any coursework or experience in GPU programming, particle swarm
optimization, or Markov modelling, I needed to teach myself these techniques. My quantitative
skills and quickness in self-study, which contributed to my success as an undergraduate BME
major at Yale, make me confident that I will be able to meet the demands of Dr. Duncan’s
research and the BME PhD program.
In my current research, I have been challenged to tackle multiple complex projects
simultaneously and independently. In only four months, I’m proud to have helped accelerate the
radiology lab’s research capabilities by helping to implement an automated segmentation tool
and by exposing a complex API for the Yale New Haven Hospital database as a GUI. Each of
these tools will save significant manual work for my colleagues and other researchers. At the
same time, I have successfully used 3D convolutional neural networks to differentiate benign and
malignant liver lesions in multiphasic abdominal MRIs, classifying six major lesion types with
unprecedented accuracy. Facing the challenge of a small and heterogeneous dataset, I developed
creative techniques to expand the training data, including generating artificial samples and
augmenting the data with nonlinear transforms.
My experience in industry has also contributed to my effectiveness as a researcher, not only
honing my programming skills but also enhancing my communication and teamwork skills. I codeveloped
software with my colleagues, regularly presented my work to a range of stakeholders,
and became comfortable managing multiple projects. As a result, I’ve been able to collaborate
effectively with other lab members and minimize distractions from my research, and I anticipate
that these skills will only become more important during my PhD.
I look forward to a career in which I can see my research help to advance medical ML and
eventually improve patient care. I feel fortunate and excited to pursue this field just as the
growing accessibility of medical imaging data and computing power is opening up numerous
opportunities for ML to transform how we diagnose and treat diseases, and I aspire to help make
this revolution in health care a reality. I believe that pursuing a PhD at Yale, and continuing my
research under Dr. Duncan would be the best step in the next stage of my journey, helping me
solidify my expertise in the field, mature as a researcher, and maximize my contributions to
computational medicine to ultimately make a meaningful impact on the quality of health care.