Teaching Philosophy

My philosophy and skillset as a science and technology educator has evolved in a dynamic environment full of exciting opportunities. I have worked on a variety of different projects, which has helped me to become a more well-rounded and versatile academic. For me, teaching is a major career aspiration that is intimately connected to academic research. My teaching philosophy is that research should drive educational lesson plans, and that educational experiences should in turn drive research strategy. Remarkably, some of my best scientific brainstorming sessions started in the classroom with my students, and ultimately turned into novel research insights. While a graduate student, I received formal pedagogical training through the Certificate in College Teaching Program at Duke University. I have been able to integrate many of the pedagogical skills acquired in this program into my teaching responsibilities. As I move forward in my academic career, I am continuously striving to develop and refine my pedagogical practice.

Teaching Experience

I currently teach MedPhys 507 Radiation Biology. In addition to these formal classes, I also teach a yearly workshop on concepts in biomedical image analysis. This 2-week workshop provides an introductory overview of practical techniques (e.g., image processing, computer programming, digital image formats, etc.), as well as current trends in biomedical computer vision applications (e.g., radiomics, deep learning, etc.).

Student Learning and Assessment

My pedagogical practice focuses on a variety of novel teaching paradigms. For example, I frequently use a technique called Visual Thinking Strategies (VTS) when I am teaching computer vision and image processing concepts. VTS - which has traditionally only been applied to liberal arts education - is widely credited with transforming the way students think and learn about complex issues. The technique is based on Abigail Housen's Theory of Aesthetic Development, and challenges students contextually, metaphorically, and philosophically through the interpretation of art. Due to its basis in visual perception and picture interpretation, I have adapted the VTS paradigm as a teaching tool for concepts in biomedical imaging, computer vision, and image processing.

In my experience, from a VTS perspective, there is a predictable manner with which students perceive and interpret biomedical imaging data. This is highly consistent with Housen's theory, and provides a reliable formative assessment of student learning. It is fascinating to see students start with a very basic qualitative description of an image, and quickly progress towards complex conversation regarding its mathematical basis. For example, VTS helps students appreciate subtle differences in image texture, anticipate the impulse response of various imaging filters, and speculate the semantic meaning of unsupervised feature extraction. Frequently, students arrive at fundamental questions, including: (a) Do we agree that we're looking at the same phenomena in this image? (b) What is the physical and biological basis responsible for such phenomena? (c) What computer vision techniques could be used to detect and characterize such phenomena?

Fig. 1. An illustrating example of a Visual Thinking Strategies exercise, “Dogs”, implemented in my Computational Imaging class. The 8 images on the perimeter of the figure are perturbed version of the image in the middle. The objective is for students to work together to in order to inversely determine which image processing technique caused the various perturbations.

Fig. 2. An illustrating example of Visual Thinking Strategies applied to medical image analysis. I often use this pedagogical technique to stimulate inclusive student discussion and critical thinking.

Fig. 3. I use simplifed visual devices while teaching to illustrate more complex technqiues, such as image texture shown in the example above.

Learning Happens in the Laboratory

It is my belief that one cannot fully disassociate teaching from research. The pedagogical methods that I employ in class have served as an interlocutor between my teaching and my research. As such, I believe that active learning in a laboratory environment is essential to prepare, educate, and mentor the future academics that will one day take our place.

Educating Physicians: The Gatekeepers of Health Data Science

A particular aspect of my job that I have always taken seriously is effective communication with physicians, i.e., the gatekeepers of health data science. I believe that one of my strongest skills as an educator is my ability to translate technical jargon into a language that non-technical healthcare professionals can appreciate and value. Many of the pedagogical techniques that I use in the classroom are equally applicable in a clinical environment, and have helped bring my operational data science projects to fruition. An illustrating example is the deep learning framework that my team has developed to segment microscopic structures on digital pathology images. This project has benefited tremendously from me teaching basic machine learning concepts to our collaborating pathologists. By teaching them enough to appreciate the technology, they have been able to more effectively communicate valuable domain knowledge.

The Future: Health Data Science Requires Interdisciplinary Education

A major future teaching goal of mine is intimately connected to a key research objective: the clinical translation of data science in healthcare. This will undoubtedly require a paradigm shift in current clinical operations, and thus new educational requirements will be needed as the field progresses. As both an imaging physicist and data scientist with over 10 years of experience embedded in a hospital environment, I am uniquely qualified to help close the educational gap between the data science community and the clinical community. In particular, I would like to teach a course open to both science/engineering students and medical students. These are the two professions that make translational health data science possible, so in my opinion it makes sense to have them train together in some capacity. In addition, undergraduate medical education would greatly benefit from having didactic data science options. This is becoming very apparent across the country, in particular due to the rise of board certification sub-specialties for physicians in clinical informatics (e.g., the Association for Pathology Informatics, the American Board of Preventative Medicine, etc.). In order for medical students to be competitive in these data-driven fellowships, data science needs to be injected into medical school curricula.