Assistant professor at Emory University’s Department of Biomedical Informatics and Department of Radiology, Dr. Imon Banerjee discusses her globetrotting education and career and reveals why nothing is ever wasted in science


Dr Banerjee is an assistant professor at the Department of Biomedical Informatics and the Department of Radiology at Emory University. Her research interest lies in the analysis and integration of unstructured medical data to support diagnosis and treatment.

Prior to joining Emory, Dr. Banerjee undertook her postdoctoral training at Stanford University under the supervision of Professor Daniel L. Rubin. She was also an instructor at Stanford AIMI; a Marie Curie European Fellow; early stage researcher at the National Research Council of Italy and The European Organization for Nuclear Research (CERN) in Switzerland.


What initially sparked your interest in medical AI?

I have always liked mathematics and computer science since I was a child but I never thought of applying them to the realm of medicine and healthcare. But while I was doing computing research after my graduation, I came across an opportunity to become a medical fellow in Italy. That was when I began my first medical AI project.

During my fellowship, I was involved in a very interesting multiscale physiological human project, funded by the European Commission. It was the building of a human atlas. Imagine a Google map, you can zoom in to a particular region. Likewise, in this human atlas, we can zoom into data at the organ level, tissue level and RNA and DNA levels.

That sparked my interest in medical AI because previously, I was just a computer scientist without medical knowledge and experience. After the fellowship, I could begin to understand the complexity of the domain and the potential of using computer science techniques to improve medicine and healthcare. So, I decided to do my postdoctoral training in that area.

I went to Stanford University and became a postdoctoral fellow turned research scientist at Professor Daniel Rubin’s Laboratory of Quantitative Imaging and AI. That officially opened my door to medical AI as I started collaborating with other researchers within and outside Stanford.

Could your career taken a different pathway?

Not really. Everything came very naturally because of my interest in mathematics, physics and computer science. However, I once thought of shifting my career pathway when I was reading my PhD. I helped out in a project with the British Museum, using computer graphics and computer vision to restore artifacts and keep them as 3D models. That was rather compelling – it’s like solving a puzzle using 3D geometry and the experience made me considered if I should pursue my career in that direction.

Coincidentally, the British Museum project also contained some medical elements because it involved the reconstructions of prehistoric animal and human bone structures and anatomy. I aborted the thought eventually and chose to focus on medical AI instead because the field was up and coming at that time and it deals with current, ongoing challenges and not something in the past.

You moved from India to Switzerland, Italy and then the US for your education and training. Was it difficult to adapt to the different cultures?

There wasn’t any huge cultural shock because scientific communities, wherever they are in the world, don’t differ much. People who are interested in science will always be interested in science regardless of where they are from and their background. This is one of the things I like about this community, it’s very tight and people are very passionate about what they do.

Who has been your biggest influence?

I always regard my supervisors as my biggest influence – wherever I’ve travelled to as part of my education and training. When I was at CERN, physicist David Francis immediately became my idol because of how innovative and supportive he was. After I received the Marie Curie European Fellowship and started my research work at the National Council of Research (CNR) in Italy, I looked up to the Research Director Michela Spagnuolo, as she is one of the few female leaders in scientific research.

You mentioned looking up to a female figure in your career. Does gender inequality in computer science and medical AI frustrate you?

Yes, it bothers me a lot. I am running a research group now and I can already spot gender inequality there because out of six PhD students I recruited, only one of them is female. So I want to overcome that by hiring more female PhD students. The problem is not many ladies are willing to invest their time in higher education.

I don’t think it has anything to do with ability or interests but the fact that many females are giving up at the very early stage of their scientific career. I think it will be great if we can educate female students early, probably at high school stage, know their thoughts and find out the challenges that are stopping them from having a scientific career.

There are signs that we are starting to see real changes though. For example, at the Emory University Department of Radiology where I am now based, we have a balanced male and female ratio of leadership figures and our Chair of Radiology and Imaging Science (Dr. Carolyn Meltzer) is female.

You mentioned many females giving up in the very early stage of their scientific careers. What motivated you to keep going?

My main motivation was my family. I count my blessing for that because I think not receiving enough support is probably another reason why women are giving up at the very early stage of their scientific careers. I was lucky to have all the support I need from my family and husband to continue my career.

What advice would you give someone starting their career in computer science or medical AI?

I always give this advice to my students. AI is an interesting and up-and-coming field of study. It’s very easy for someone to find data and codes that they need to build a model, without having an in-depth knowledge or foundation for the subject. The trend is upsetting since it will not lead to real innovations. It can also be dangerous especially if a risky or controversial subject like medicine is involved.

What one ought to do is to understand what AI is, what AI can do, its limitations and so on. Next, have a thorough understanding or work with someone who knows the field that the AI model is intended for. AI is larger and more complex than model building, model development and model training.

What was the best piece of advice you ever received?

That came from my supervisor at Stanford, Daniel Rubin. At Stanford, there are a lot of opportunities to apply for research grants for someone to start their research projects. Professor Rubin was the one who encouraged me right from the beginning to apply for my research grants because he knew that I wanted to stay in academia and be a medical AI researcher. Yet, the acceptance rate, especially for some of the popular funding programs, is about 5% so I was quite disappointed as it was just so challenging to apply for my research grant.

But Professor Rubin told me, never throw away any piece of innovation. Nothing should be wasted in science. If you are interested in something and you discover an innovative solution there, you can reuse it anywhere. Never get disappointed when your innovation gets rejected, use it somewhere else. That was great advice. It’s always going to be challenging for anyone starting their career in academia but the key is, never give up.

What’s left to conquer?

I am very interested in fusion model. At the moment, medical AI is still relatively imaging-focused. I don’t think having imaging data is sufficient for any diagnosis or prediction. If we wish to deploy AI in real clinical settings or support clinicians in making life-changing decisions, we need more different types of data from various sources.

Ironically, clinicians are often overwhelmed by staggering amounts of heterogeneous data. I think AI can play a big role in creating a digested version of the clinical information so that clinicians won’t have to spend time searching for the piece of detail they need. I am very interested in making such a model and it’s a direction that I intend to pursue.