Questions and Answers

1. What was the major objective of your project and what was your plan to achieve it? 

The major objective of my project was to determine which commonly used signal analysis method, between the fast Fourier transform and the continuous wavelet transform, was superior in detecting the presence of Major Depressive Disorder. My plan was to use both transforms to analyze a publicly available EEG dataset of a patient with Major Depressive Disorder and ultimately output a resultant topographic map of the patient’s brain for each transform. Then, I would assess each topographic map for the strength with which it presented depression biomarkers. The topographic map that confirmed the patient’s depression more strongly would correlate to the better analysis method for this purpose.

       a. Was that goal the result of any specific situation, experience, or problem you encountered?  

This goal was a result of a lifelong awareness of mental health issues in my community, especially among close friends. After I began brainstorming ideas for my research with mental health on my mind, I remembered a friend who told me that they purposely minimized the extent of their symptoms when filling out the diagnostic surveys doctors use to diagnose depression. They did so because their parents were uncomfortable with the stigma of mental health issues. I was concerned that there were others who also, by virtue of this somewhat subjective diagnostic process, were not getting the help they needed. As such, I began to look into the surprisingly sparse field of quantitative depression diagnosis. I was intrigued by the analysis of EEG data, which is a rich, complex form of neurological data, and how Fourier transforms, which I’d been discussing with my mentor, could apply in the context of depression diagnosis.

       b. Were you trying to solve a problem, answer a question, or test a hypothesis?

While I did not design my research to solve the broader problem I was concerned in, which was quantitative depression diagnosis, I hoped that the question I answered — what signal analysis method is optimal for depression diagnosis? — was a step in the direction of a general solution or invention.

2. What were the major tasks you had to perform in order to complete your project?

- I read through existing literature, conjectured a hypothesis, and designed the experiment

- I learned the necessary MATLAB programming language in just a few weeks. Most importantly, I learned how to pre-process/analyze EEG data and output topographic plots in MATLAB using both types of transforms with the FieldTrip toolbox

- I learned significant non-Fourier transform background knowledge (e.g. the nature of EEG data collection and analysis, what wavelet transforms are and how they mathematically and practically differ from Fourier transforms)

- My data analysis and the conclusions I drew were also drawn completely on my own (with the help of pre-existing literature, of course)

-I wrote my whole paper and created most of my diagrams on my own

       a. For teams, describe what each member worked on.

 N/A

3. What is new or novel about your project?

The concept and goal of my project are novel. Quantitative depression diagnosis in general is a surprisingly under-researched aspect of neurological data. My methods, too, are novel in the specific way I implement them.

       a. Is there some aspect of your project's objective, or how you achieved it that you haven't done before?

Yes, there are many aspects of this project that are new to me, as this my first formal research project. As I’ve articulated in other parts of the answer section, the code and processes by which I reached my conclusion were all novel to me; I had to learn them over the course of my research. Although I had preexisting knowledge of Fourier transforms through conversation with my mentor, I had never used them to analyze EEG data. I also had to apply wavelet transforms, another novel aspect, among others.

       b. Is your project's objective, or the way you implemented it, different from anything you have seen?

Yes! I haven’t seen any studies focusing on the analysis of EEG data with respect to quantitative depression diagnosis. I also have not seen any studies that have used topological maps of the brain as a parameter by which to measure the effectiveness of a signal processing technique for neurological data.

       c. If you believe your work to be unique in some way, what research have you done to confirm that it is?

I’ve done an in-depth literature review on the above points. I’ve also discussed my research informally with parents’ friends who are medical professionals in the neuroscience field. 

4. What was the most challenging part of completing your project?

My biggest challenge in carrying out my procedure was in learning to write the necessary code to take in EEG data and output topological maps with both transforms. Though I’d hoped for my mentor’s familiarity with the MATLAB language in his field to assist me, he unfortunately could not aid with my desired analysis, as EEG data was very different from what he worked with. Ultimately, though, it makes me proud to look back on my code, knowing the many hours of hard work those lines hold.

      a. What problems did you encounter, and how did you overcome them?

- I struggled to choose how I should analyze the EEG data, especially given its complexity. I didn’t have the extensive background needed to interpret the rawest form of the transformed data, which looked like a spectrum of many different colors and imperfections. Through my research, as I watched YouTube video lectures to help me understand the nature of EEG data and the MATLAB language, I found a resource: the FieldTrip toolbox. The FieldTrip toolbox was an additional extension that allowed me to create topological maps of the brain, which would not be as difficult to interpret. Through helpful articles on the FieldTrip website, many StackExchange threads, and the FieldTrip and MATLAB documentations, I succeeded!

- As I researched, I realized the different units by which the transforms expressed intensity (power vs. scale, respectively) would make a quantifiable comparison difficult. As a result, I also learned how to normalize data and adjust units of the CWT output to match those of the FFT. While this did not completely eliminate the issue, it did reduce it.

      b. What did you learn from overcoming these problems?

On a concrete level, I gained a lot of technical abilities — MATLAB programming skill, EEG data analysis, and knowledge about Fourier transforms and wavelet transforms. More importantly, though, I learned how much I enjoy the problem-solving identity of research. I love this process of learning a lot about a subject in a very short amount of time; I love going deep into a slice of something much bigger. As someone who loves exploring different areas, it's refreshing to know I have the capacity to go this deep into one and to be thrilled at the prospect of going deeper still. 

I also reaffirmed the knowledge that I love challenging myself. Though frustrating at times, I loved learning MATLAB at questionable hours in the a.m., and I loved applying

Fourier transforms in a biological application (as opposed to the physics applications my mentor’s area of expertise was) simply because I really wanted to research something related to depression. I especially loved that the research I did felt meaningful. Depression is something that has thankfully not afflicted me, but I've always been concerned for the mental health of those around me. 

5. If you were going to do this project again, are there any things you would do differently the next time?

If I were to redo this project, I would broaden the scope of my data. Specifically, I would compare the topographic maps of both healthy and unhealthy (depressed) subjects, to control for natural differences in the way the Fourier transform and wavelet transform analyze certain frequency ranges. I would also aim to compare more types of signal analysis methods from the start, as I realized later on in the project that my code could potentially work very well with other common transforms, with some adjustment. 

6. Did working on this project give you any ideas for other projects? 

Yes! After working on this project, I found myself intrigued by the types of data used in neurological studies. EEG is one of the most common, but there are others, including functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and single-photon emission computed tomography (SPECT). I’d like to compare these data types on the same set of subjects and with the same analysis methods, to understand how each translates to depression detection. 

In the future, when I have more resources and knowledge at my disposal, I’d also like to look into machine learning and generative AI algorithms for EEG data. Our quickly-improving age of information processing aligns very well with the information-rich, but often confusing, nature of EEG data. This is an area I’m actually very excited to pursue in the future!

While not a formal research interest, I’m actively learning about the biological mechanisms of depression (such as tissue decay in certain areas of the brain), as well as how the different frequency bands used to analyze EEG data have been historically determined.

7. How did COVID-19 affect the completion of your project?

N/A