Nisha is a student from the United States who attended the Summer STEM Institute (SSI). Nisha is interested in the application of technology to healthcare. Before SSI, Nisha interned at Cisco and also participated in math and computer science olympiads, qualifying for the American Invitational Mathematics Examination (AIME). Over the summer, Nisha worked on a project, “Interpretability for the Automated Diagnosis of Musculoskeletal Radiographs,” to increase the transparency of machine learning algorithms used for musculoskeletal disease diagnosis. Nisha’s paper was accepted to the Harvard International Young Researchers' Conference, and you can read her full paper here.
I think my love for science started with Khan Academy. I remember that I used to binge-watch their videos as a fifth-grader, and I was obsessed with their series on how the heart works. After that, I continued building upon my love for science and took science classes at school. I had inspiring teachers, and my friend group has also always been pretty science-oriented. One of my friends started doing research, and he was really into it. He suggested I should try it out because he thought I would really enjoy it too, and that’s how I became interested in research specifically.
No, SSI was actually an introduction to research for me. I had no idea what the research process was. I did come in with a programming background though. I took data structures and algorithms at Stanford Online High School and participated in the USA Computing Olympiad (USACO). I knew how to program, but I knew nothing about machine learning. I read articles about it and thought machine learning was really cool, but I didn’t really know how to do it myself.
Well, competitive programming is much more algorithmic and theoretical. You think about things like how to solve problems efficiently, analyze data structures, and construct elaborate proofs for the time complexities of algorithms. In data science, there’s so much more leeway and you get to decide how to represent and interpret data. A huge part of data science research is not just coding, but knowing where to find resources. Before SSI, I had read articles about the tremendous impact of technology in the real world, I had never been directly involved with its application.
To be honest, I was never super interested in technology as a child. I thought I was going to become a cardiothoracic surgeon when I grew up, and I was really interested in medicine. In my childhood, I had a lot of experiences as a patient in a hospital, so doctors were my role models. As I grew older, I realized that I can reach a lot more people through technology. The scope of people you can reach if you make something is much higher, and that’s why I wanted to dive into technology.
I think CRISPR-Cas9 research is really interesting. I like reading about technologies like genetic engineering. I like thinking about how technologies are distributed throughout society and their accessibility to people in different classes. I recently read Brave New World by Aldous Huxley where characters live in a technology-oriented society and all humans are engineered to be identical. The society takes out the individualism of the human experience to make society functional and stable, and I found the book very fascinating. It’s interesting to think about how technology can be used irresponsibly by governments.
I’m really involved in Taekwondo. I used to train a lot when I was younger, but I’m actually a coach now. One of my favorite things to do is coaching younger students. Beyond Taekwondo, I’m also a part of choir. The choir community is a really big deal to me. I obviously love to sing, and there’s just something special about singing amongst other musicians. It’s a different type of dynamic. I especially love show choir, which is like musicals. It’s so much fun when we do stuff from Disney or Broadway musicals. There’s all this dancing, singing, and stunting on stage.
Outside of Taekwondo and choir, I really enjoy calligraphy. I’ve been doing it for a while, and I love exploring hand lettering. I also love dancing. It’s a huge stress reliever for me, and I can just spend hours trying to learn dance routines.
Sure. Although machine learning models that are applied to healthcare have significantly improved in performance, these algorithms are often seen as “black boxes” because it’s difficult to understand how or why a model makes the predictions it does. My project looked at musculoskeletal x-ray images and diagnosed disorders within them while increasing the interpretability of models to increase the transparency and trust within these algorithms. For example, my model uses saliency imaging to highlight which regions of an image were abnormal or indicated disorders. Ultimately, the goal of the project was to help gain patient and physician trust when it comes to applying machine learning models to diagnosing musculoskeletal diseases.
I wanted to do something that involved both medicine and technology. After conducting a lot of background research, I started wondering why AI isn’t used more prominently in the healthcare system. While reading different articles, I learned that many physicians don’t view AI as trustworthy because it’s not explainable. This inspired me to dive further into interpretability and healthcare.
First of all, I think I became so much better at searching for things on Google and scoping good articles to read. Before, I never searched up this many questions or bugs. Also, before, I had no idea how to conduct research, so I was able to go through the whole process for the first time. It was really, really challenging, but I had so much support around me. I feel like you never really know how to do research until you do it yourself. Also, of course, I was introduced to machine learning and data science. I learned so many interesting perspectives that I will carry with me throughout my educational career.
My favorite part was getting the results. When my model was training, I was very anxious while waiting for the results. I was so excited to see the results, and after seeing them, I learned so much about how to represent results. I learned all about saliency curves and how to visualize results.
Something really interesting I learned was the way we think about accuracy and why it’s not a good metric. I never really thought about it before until this summer. During a research lecture, I learned about how you can have a data split of say, 95% positive and 5% negative. Then, if you guess positive for all of them, you’ll have a 95% accuracy, even though your algorithm is very naive. This was a lightbulb moment for me.
My favorite part about SSI was definitely the Masterclass series. It was just so amazing, and I got to meet all these really cool people and learn from their high school experiences. I remember that there was this one Masterclass on time management, and the speaker had super intricate graphs about how they went about their day. I thought it was so clever and was in awe of the amount of self-discipline it took. It was really inspiring to hear how others prioritize their time and use their abilities and talents.
I’ve definitely tried to improve it to the best of my ability. It was obviously different without the SSI community and daily meetings, but I think the groundwork had been laid down to a point where I felt like I could make improvements by myself. I added line classifiers, which is an imaging technique, which was really cool.
I was inspired by the email prompts I learned about at SSI and reached out to professors through those. I’ve recently started a new project on pulmonary embolisms at Stony Brook University that has been really interesting. We’re working to diagnose pulmonary embolisms from CT and Q scans because it’s really easy to miss these particular pathologies. Also, I received the opportunity to work on Ciona intestinalis research at UC Barbara. We’re looking at Ciona intestinalis organisms and their neural networks to use computer vision to recognize synapses. SSI definitely opened up a lot of opportunities for me.
My research project made me really aware of the fact that I want to do undergraduate research when I go to college. It’s not something I thought about before, but now, it’s insistent on my mind. When I go to college, I’m definitely going to continue researching. Also, throughout high school, I always thought I would study computer science, so that didn’t really change. However, SSI solidified my passion for computer science, and now I’m sure that I want to continue using computer science for applications in the healthcare system and beyond.
I would love to become a computer scientist and engineer. It’s really exciting for me to work on machines, and I enjoy working on technology that can help causes that really matter to me. More specifically, I would like to help make healthcare more affordable and accessible. I’m particularly passionate about providing women with access to sanitary products. I recently read an article about someone who is making really affordable pads for women in poverty areas. I think it would be really cool to work on an issue as impactful as this.
Outside of work, I’ve always wanted to make videos for Khan Academy because so much of my educational experiences were shaped by them. Also, writing a book has always been on my bucket list. I love traveling, and I would love to write a narrative on my experiences in different cultures. Different cultures around the world have always fascinated me, and I would love to merge my interests in travel and narrative writing.