From Lab Experience to Original Research: Lukas’s Journey in Biomedical ML

Lukas investigated how signal quality affects deep-learning models for electrocardiogram (ECG) classification using PTB-XL, a dataset of 21,837 clinical 12-lead ECG recordings. Through experiments spanning five diagnostic classes and ten independent training seeds, he identified class-specific effects of signal quality and a loss-reweighting normalization pitfall that can unintentionally weaken learning for other diagnostic classes.

The project developed through substantial experimentation. After testing eight model and training interventions that did not produce the meaningful improvement he was seeking, Lukas shifted his focus toward understanding the training data itself. His research examined how different forms of signal interference affected particular diagnostic categories and why an attempted correction could introduce a separate problem elsewhere in the model.

Lukas entered Echelon Scholars with experience from a local wet-lab internship and a desire to strengthen his technical research abilities. Over several months, he progressed through literature analysis, methodology development, repeated experiments, statistical evaluation, and scientific writing. With detailed mentor feedback, he refined both the research and its claims, checking alternative explanations and developing the findings into a complete postgraduate-level paper.

In the Scholar’s Words

What originally drew you to Echelon Scholars?

“I had previously completed an internship at a local wet lab, which gave me some experience with research. When I came across Echelon Scholars, I saw an opportunity to take the technical side of my research further and work through a more advanced process to develop my own paper.”

What was the most interesting part of developing your research topic?

“One of the most interesting parts was near the beginning, when I was reviewing the state-of-the-art research and narrowing down my topic. I enjoyed looking at the limitations of existing methods and bringing those ideas together into a focused research question. It helped me move from a broad interest toward something specific that I could investigate.”

What was it like moving from reading research to building something yourself?

“The technical work felt daunting at first. There was a lot of code, and I had to think about how to build something from the ground up. It was helpful to start with the state-of-the-art methods, understand how they worked, and then work toward creating something novel myself.”

What was the most challenging part of the project?

“The methodology stage was probably the most challenging. I had an initial idea of what I wanted to investigate, but after a lot of technical iteration, nothing seemed to produce a meaningful improvement over the existing methods. I had to pivot several times and rethink where the novelty could come from. Finding a direction that made a real contribution took a lot of work.”

What did you value most about working with your mentor?

“I really appreciated how quick, responsive, and detailed my mentor was. When I had questions, he replied quickly and gave thorough feedback. I also liked the process itself because it pushed me to elevate my own work and develop the paper to a higher standard.”

How did your technical and statistical abilities develop?

“I learned a lot about machine learning and statistics. In particular, I became much more comfortable understanding statistical significance, evaluating different metrics, and interpreting the results. Those details became an important part of understanding what my experiments were actually showing.”

How did the experience improve your scientific writing?

“My writing definitely improved. Reading research papers helped me better understand how researchers structure their arguments and explain their findings. As I worked on my own paper, I became better at presenting the methods and results clearly and bringing everything together into a complete research paper.”

What did completing a postgraduate-level paper mean to you as a high school student?

“It felt like a very big accomplishment. Looking back at where I started, I would not have imagined being able to produce the final paper. Over the months we worked together, I really elevated my work. Seeing the difference between the beginning and the finished paper made me realize how much I had learned.”

How will you build on this experience in your future academic work?

“I definitely want to continue doing research. This experience has given me a stronger understanding of the technical fundamentals, how to read research papers, and how to write my own. I think that will make the process smoother when I explore other topics and fields. A lot of what you learn is in the nuances that you would never understand just by looking at research from the outside.”

What would you tell another student considering Echelon Scholars?

“I would recommend joining, especially if you want to challenge yourself and develop your research abilities. It is a long process, and much of it is difficult, but going through that rigorous journey makes the outcome meaningful. You develop a much deeper understanding of the work, and having the option to receive college credit is an additional benefit.”

‍

"Through Mr. Kulkarni's guidance, I learned how to read a paper: splitting it into parts and starting with the most digestible ones, the abstract and discussion sections. This helped me with my research proposal and with school and research generally. A graduate student at my internship told me it was the hardest thing she had to teach herself in college, so I'm glad to have learned it in high school. Mr. Kulkarni let me work through the research on my own, which helped me build research skills, and his weekly updates and follow-ups helped when I got stuck. Through many revisions, what once sounded like a reach became a paper we submitted to IEEE DLB2H."