The applications of signal processing in the field of healthcare are just mindblowing, thanks to advancements in machine learning technologies.
Shiv Gehlot, our next pathbreaker, Lead Engineer (AI) at AIRA MATRIX, works at the intersection of deep learning and medical image analysis, with the aim of diagnosing underlying medical conditions using AI and medical images.
Shiv talks to Shyam Krishnamurthy from The Interview Portal about his PhD on Deep Learning Assisted Medical Image Analysis which was focused on computer-aided blood cancer diagnosis using microscopic images.
For students, the rise of sophisticated computational technologies have brought exciting career opportunities at the intersection of signal processing and AI !
Shiv, tell us what were your early years like?
My native place is a small village in Uttar Pradesh where farming was my parents’ primary occupation. I studied there till high school. Later, I moved to Delhi for the rest of my schooling and higher education. In conjunction with the prevalent culture in India, my family too had engineers, and they had a strong (positive) effect on me. Naturally, my parents also wanted me to be an engineer, but I also seemed to love this idea since childhood, long before I was even aware of the formal procedure to get into engineering. This early goal setting helped me in making the relevant decision based on my interests rather than trial and error. Hence, I went for engineering because it was my selection and not because “everyone else is also doing it”, which I believe could have a severe negative impact on students, having witnessed it myself.
What did you do for graduation/post graduation?
Within engineering, electronics excited me the most as I have seen people working with circuits. It amazed me how a microcontroller, along with other components and a few lines of code, could do wonders. Even the basic circuits, such as rectifiers, aroused curiosity in me, and I wanted to delve into this area further. Hence, I opted for Electronics & Communication Engineering (ECE) for my B.Tech and never regretted a day. The courses were exciting and fun to learn.
In ECE, there are many different categories of courses focusing on electronics, communication, and signal processing, to name a few. The signal processing field stood apart to me, and I enrolled for an M.Tech degree in Signal Processing. I subsequently did my PhD in Deep Learning Assisted Medical Image Analysis from Indraprastha Institute of Information Technology, Delhi.
What were some of the events that influenced you to take up a career in Medical Imaging?
I have always wanted to go for engineering since childhood. However, once I started engineering, the decisions were influenced by my learnings, experiences, and interests. For example, in B.Tech, I developed an interest in M.Tech, and while in M.Tech, I realized my interest in research, so I went for PhD. Similarly, the field of study was decided based on my recent interests. For my M.Tech, I chose Signal Processing based on my interest in the courses in B.Tech. For my PhD, I wanted to work on image processing, and as deep learning was really catching on in 2017, I decided to do research at the intersection of deep learning and medical image analysis.
How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path
I enrolled for M.Tech through the Graduate Aptitude Test in Engineering (GATE) from Netaji Subhas Institute of Technology (NSIT), University of Delhi. Hence, I obtained the scholarship from the MHRD as a monthly stipend. After M.Tech, I wanted to enroll for PhD, but a few other responsibilities took over. Also, my M.Tech. was in continuation after B.Tech, without any drop year, which students usually take for the GATE preparation. I appeared for GATE based on self-study only and got a decent rank to get into NSIT.
Hence, I wanted to cool-off before enrolling for a PhD, because I knew that a PhD is a long-term commitment and will be very exhausting. I chose academia for my first job as I thought it would be easier to switch to PhD as compared to the industry due to the similarity between the two domains. Hence, I started working as an Assistant Professor in the ECE department at an engineering college. At first, handling a class of 60+ students was challenging, and honestly, it was chaos. Though I had all the required knowledge, disbursing that knowledge is an art. Teaching is an art. People with mediocre knowledge can be exceptional teachers, and people with exceptional knowledge can be mediocre teachers. Eventually, I developed the required skills along with some management skills. Nevertheless, I realized this was not the profession for me as I really did not enjoy it. This realization was helpful as it paved the way for my post-PhD plans. I decided not to go for academia after my PhD and that helped me focus on my end goal during my PhD.
For a PhD, you require a source of funding. The institute can provide it, or there can be some external fellowships, typically more generous, with some other added advantages compared to the institutes’ fellowship. Through the National Eligibility Test (NET), I eyed the UGC Junior/Senior research fellowship provided by Govt of India. One can appear for the NET after post-graduation in the required stream. The selection is very competitive, and I secured the fellowship. The next step was the selection of an institute and an advisor, both of which are very crucial. While for B.Tech or M.Tech, a good college will suffice, for PhD, a good advisor is equally necessary. Based on my research interests, faculties, and the facilities available, I enrolled for the PhD in IIIT-Delhi, which was one of my best decisions. IIIT-Delhi stands out in terms of providing a tremendous global research environment while equally focused on the fantastic facilities provided to the students. It is a fast-moving environment with everything available 24*7.
As with every other PhD student, I spent most of my time in the lab, including weekends. PhD life is very complex, with so many ups and downs, and it is impossible to capture it in a few sentences. To summarize, I worked at the intersection of deep learning and medical image analysis with publications at some of the best venues and international internships.
As I had already decided that I would not be going for an academic career, I started looking for an industry job mainly in the same domain as my PhD research work after completing my research. However, I kept myself open to other areas. Eventually, I had multiple offers with job descriptions related to my research work and different from it.
During my PhD, a significant part of my research was focused on computer-aided blood cancer diagnosis using microscopic images. Specifically, I developed a deep learning-based methodology for diagnosing Acute Lymphoblastic leukemia (ALL) and Multiple Myeloma (MM). Apart from microscopic images, I also worked on other modalities, such as whole slides images and DICOM images. From an applications perspectives, I have worked on classification, segmentation, and stain normalization utilizing supervised and self-supervised approaches. Apart from research, I was on the organizing committee of two international challenges organized as a part of The IEEE International Symposium on Biomedical Imaging (ISBI) 2019 and 2021, focusing on classification and segmentation, respectively.
Another part of my research was on the reconstruction of compressively sensed data. With this approach, I explored the imputation of gene expression data and reconstruction of hyperspectral images.
During my PhD, I spent one semester as a research intern at the College of Health Solutions, Arizona State University.
My list of publications are available at:
How did you get your first break?
I have relied on networking for job search. It is necessary to have a good profile, but it is equally important to have a good network. LinkedIn is powerful and can be beneficial in job search; I have received multiple offers through it.
What were some of the challenges you faced? How did you address them?
Challenges are a part of our life, we evolve by overcoming them. Every challenge is unique in its own way, and so is the strategy to counter it. However, I would like to broadly classify the challenges in two categories, both related to decision making.
- The comfort zone challenge.
A Comfort zone is a situation where there is no or minor turbulence at most. The routine is fixed, and you are in control. People usually get hinged to these conditions, which are difficult to come out of. Unfortunately, in such cases there is no longer a learning curve due to lack of new challenges, which eventually blocks career progression. It is very challenging to address such situations due to denial and fear of losing control. Even after acknowledgement, it is not easy to take steps as it disturbs our joyous routine. I believe acknowledging and addressing the comfort zone challenge is crucial for a good career trajectory.
2. Challenge to decide between immediate gain vs. long-term gains
It is tempting to choose immediate gains rather than committing to a long-term goal. Setting a long-term goal usually requires dedication and hard work, and it comes with new challenges. Hence, it always seems beneficial to pick the immediate rewards. It is not always the case. It is always an excellent strategy to weigh the pros and cons of both situations and decide the better path irrespective of the challenges and situations.
I have faced severe financial, emotional, and health breakdown phases in my life, so much so that life was nearly impossible during those phases, and I was in the dark. However, I never gave up and always followed my heart and the above two principles, and eventually witnessed better days.
Where do you work now? What problems do you solve?
I currently work at the AIRA MATRIX as a Lead Engineer (AI). My present work is in the same domain as my PhD research work. My work is at the intersection of deep learning and medical image analysis. To summarize, my work is centered around diagnosing underlying medical conditions using AI and medical images. A simple and relevant example can be the diagnosis of COVID-19 using chest X-rays with the help of AI.
As my PhD research work was in the same domain, I developed the required skills during my PhD. However, deep learning is a rapidly growing field with hundreds of research papers being published every day. Hence, I rely on relevant and important research papers apart from textbooks or online lectures to keep my skills updated.
My work involves understanding the problem statement and developing a methodology from theory to implementation in order to obtain the required performance and results. As I am very passionate about this domain and research, I enjoy every bit of my work irrespective of the challenges.
How does your work benefit society?
Deep learning, or machine learning, has been the fastest growing field over the last 5-6 years. This field has penetrated every domain and found immense applications. You are already surrounded by its applications, from the smartphone, autonomous driving, finance to the medical domain, to name a few. Hence, it brings exciting career opportunities and is the right time to step in this direction.
The taboo around a PhD is also gradually fading away. Earlier it was believed that a PhD was required only for academics. However, if you acquire the right skills during the PhD, and focus on practical applications apart from theoretical knowledge, then there are sufficient industry opportunities as well. You have options to choose between an academics or industry career after PhD. Candidates interested in academics usually go for Post-Doc first. In India, the research scenario is improving significantly, with many institutions focusing on expanding and enhancing their research facilities.
Tell us an example of a specific memorable work you did that is very close to you!
I am passionate about my work and do it with a whole heart. It is tough to choose one over the other. Every task comes with challenges and requires a different thought process and methodology. However, my first few research papers will always be memorable. It takes significant effort to publish your first few papers as you are trying to cross a threshold that validates your research capability. It gives you confidence and ensures that you are in the right direction. I believe my best work is yet to come, and when it comes, there will be some other benchmark, and the process continues.
Your advice to students based on your experience?
“Stay Hungry Stay Foolish,” Steve Jobs
Never stop being a student, no matter where you are. The day you start feeling more intelligent than everyone else, your growth is already stalled, and you will witness decline in the long term. Always be hungry and curious to know more, and keep the learning curve intact.
“If you can’t fly then run, if you can’t run then walk, if you can’t walk then crawl, but whatever you do you have to keep moving forward,” Martin Luther King Jr.
Life will not always be generous and straightforward. You may face unseen obstacles that may deter your courage. However, the trick is to keep moving, irrespective of the situations and challenges. Break your main goal into smaller goals, and keep chasing them.
I have an immense interest in Signal Processing and AI. All of my work has been centered around these two domains. I look forward to doing some impactful work and growing in these domains.