2024-08-01: My PhD Voyage
One day, as I sat playing Candy Crush for the hundredth time on the sofa, I thought, "I could be doing something more productive with my spare time." While there were probably hundreds of other things I could have considered, this made me start contemplating my third return to college to pursue a doctoral degree. A pretty decent tuition assistance program provided by my employer made the thought even more tempting. Therefore, I decided to proceed with submitting an application for graduate school. I also began looking into ongoing research at Old Dominion University (ODU) and contacted some professors as potential advisors.
I was at Disney World in Florida with my family in August 2016 when I took a pause from the day's activities to check email on my iPhone. At that time, I saw the email informing me that I had been accepted into the Computer Science PhD program at ODU. Although I applied earlier in the year, some recent unfortunate events made me question whether to move forward. However, refusing to be defeated, I proceeded with signing up for classes just prior to their start.
Humble Beginnings
When I first started the PhD program, I signed up for a regular course, Introduction to Bioinformatics, and a graduate seminar course in which I started some initial research and investigation of static code analysis frameworks, mesh generators, and triangulators. They served as a good transition into the doctoral degree program and are interesting topics. I had used static code analysis tools at work numerous times but never gave much thought to the science behind them. I also learned much more about chromosomes and DNA than I expected in the Bioinformatics class. I even bring up some of the things I learned when family or friends discuss their Ancestry DNA results.
With working a full-time job and having a family, I tried to make sure I found a balance between everything. Therefore, I would try to study when I had random pockets of free time, like in the bleachers for the duration of my son's basketball practices and later in the evening when most of the family had settled in their rooms to relax or watch television. I decided to take a break from classes in the summer after my first two semesters in the program so as to not overwhelm myself.
Winds of Change
Computer science is very broad, encompassing programming, operating systems, networking, cybersecurity, interface design, cloud computing, big data, artificial intelligence, and more. Being a software developer for one project for several years at a time did not necessarily give me exposure to the latest trends in some of these areas that could be explored. As I started taking different computer science courses to meet the core course requirements for my degree program, I became more interested in the topics covered in these classes. The Information Visualization course I took had an extra connection with my artistic side. In the fall of 2017, I particularly remember driving my car early and late in the day with an on-board diagnostic (ODB) device plugged into it and my mobile phone mounted on the steering wheel while wearing a smartwatch to collect data for a distracted driving project for my Mobile Computing class. That was probably the most enjoyment I have gotten out of collecting data. We also read several papers in that class related to emerging research that were very interesting.
I had continued my research in triangulations and formal methods. However, in early 2019, I decided to go with my gut and explore other research opportunities. I revisited some of the research being led by faculty members in the computer science department. I had also learned of Dr. Sampath Jayarathna, who had only been with ODU for one year at the time, and was leading research in machine learning applications. I decided to reach out to him to learn more. We discussed investigating the use of machine learning to classify electroencephalography (EEG) data captured from control subjects and those with Post-Traumatic Stress Disorder (PTSD). The topic was very captivating and innovative, and by the beginning of May 2019, I had joined the Neuro-Information Retrieval and Data Science (NIRDS) lab.
I started researching machine learning, automated medical assessments, and EEG signatures for PTSD. Soon after, I wrote a survey paper on automated PTSD diagnosis, which I submitted to the 2019 IEEE IRI Conference. The paper was accepted, so at the end of July 2019, I flew out to Los Angeles to present what would be my first publication. I must say, it was pretty exciting, and I had never been to LA. The NIRDS lab had also acquired the funds to purchase a very nice mobile EEG headset. I was looking forward to trying it out and using it to support further research once it arrived.
Life's Challenges
Originally, when I started the PhD program, I had intended to eventually apply for a program through my employer that would allow me to reduce the number of hours I worked so I could focus more on school. However, I had taken a new promotional position a couple of years into the PhD program, which would prevent me from doing so. Therefore, I had to continue my balancing act between a full-time job, family, and school.
Also, as we all know, 2020 introduced us to COVID, which brought with it mandatory telework, face masks, isolation guidance, travel restrictions, and all sorts of other changes to daily life. Personally, those changes also included a third child for me. I'm not sure when the others in my research group realized it since all they ever saw was my face on Zoom meetings, and my son was born during the school's winter break later that year. I guess COVID saved me from walking around campus with a large belly and swollen feet.
However, the pandemic also prevented me from using that fancy new EEG headset our research lab had obtained. Faculty expressed concerns about the challenge of obtaining IRB approval for research involving human subjects during COVID-19. Even the feedback I received at the 2020 AIME Doctoral Consortium included a recommendation to find existing data to use. Therefore, at that point, I shifted gears and started my quest to find preexisting EEG datasets to support further analysis.
Finding Inspiration Anytime, Anywhere
I came across different datasets on the Internet, like the TUH EEG Corpus, TDBrain, and those found in the PhysioNet repository. I was also able to obtain some PTSD related EEG datasets from EVMS as well as some data previously collected for other in-house research. Unfortunately, I found that all of the datasets were in different formats, included different metadata, and were collected using different research protocols. The tools that everyone seemed to be using to read and manipulate EEG data were not the most intuitive, although some had more promise than others from my standpoint. I also obtained some Python code, but it had to be modified to even read the datasets I was working with. After researching and reading several technical blogs and message boards, I realized a lot of other researchers shared my EEG data challenges.
However, one day, I was speaking with a guy who worked for AWS. I told him about my research and the EEG data collection, and he asked one question that stuck with me: "How much data is that"? I did not think to ask him what he was thinking at the time, but I later began to wonder how I could leverage AWS to alleviate users' challenges that made storing, reading, and preprocessing EEG data so complex. I established a free AWS account and was inspired to construct a solution that utilized several AWS cloud services. I got up and running fairly quickly through the help of multiple online tutorials. I also discovered the Registry of Open Data on AWS, which includes the OpenNeuro data repository. This allowed me to incorporate direct programmatic access to data in the system, thereby avoiding time-consuming data downloads only to re-upload data to another system.
Downhill from Here?
When I finally came up for air, I had developed a new cloud-based software-as-as-service (SaaS) solution for EEG data preprocessing that could handle a large amount of data, read different data formats, and directly pull data from cloud-based repositories. I eventually gave the system a name: Microservices for Innovative Neurotechnology Development in the Cloud (MIND Cloud). The system, which abstracts the complexities of existing pipelines and software solutions, seemed to make sense. However, I needed to demonstrate that the system actually improved the user experience as a more usable approach. Therefore, I collected performance measurements for it as well as for a baseline system, which drove some improvements to the system. I also drafted IRB paperwork for a user study that was eventually conducted to evaluate the system. Through the study, MIND Cloud was shown to reduce task time and errors. It also had a high level of usability and reduced cognitive load based on the System Usability Scale (SUS) and NASA Task-Load Index (TLX) questionnaire results, respectively.
Ending Another Chapter
I had already participated in the 2024 ODU spring graduation ceremony since there is not one during the summer months. After being in the program for so long, it seems strange to not have an assignment or paper waiting for me. I also have several people at work and in my family who started referring to me as a doctor. Looking back, it was a big commitment but a big accomplishment. However, now that my youngest son is three, I'm sure I will use the free time I have gained to support preschool lessons and more outdoor activities. I also told one of my sisters that I would try one of her afternoon naps when I get a chance.
--Bathsheba (@sheissheba)
Comments
Post a Comment