This summer, I had the opportunity to intern for eight weeks for Seraphina Solders, a graduate student at the University of California San Diego (UCSD) in the Neurosciences Graduate Program. Seraphina is investigating a new theory that Alzheimer’s Disease (AD) begins in the Locus Coeruleus (LC). She is working with a specific type of Magnetic Resonance Imaging (MRI) called diffusion MRI that tracks the diffusion of hydrogen protons in water molecules across the brain in order to map the white matter pathways.
Before officially starting to work for Seraphina, I read a few of her blogs that are published on Neuwrite San Diego. I began to familiarize myself with where AD begins in the brain, how diffusion MRI works, and how magnets are used to capture images of the brain. I also read a scientific review paper that focused on how the LC and LC neuropathology relates to cognitive function and aging.
My first project was to complete a literature search. I was looking at review and experimental papers from 2010-2021 on PubMed, a public database that gives anyone access to biomedical and life sciences literature, to investigate what tools and software researchers were using when taking diffusion MRI scans. I found 5 papers from each year in my search. I learned a lot about how research papers are structured and found the tools and software usually listed in the methods section. However, diffusion MRI has many other names, including Diffusion-Weighted Imaging (DWI), structural connectome imaging, Diffusion Tensor Imaging (DTI), and more, which made the search challenging at times. Here is a sample of the papers I found from my literature search:
For my second project, I worked with RStudio, the environment for R, which is a programming language. I took coding primers in RStudio to learn how to utilize data science coding. In one of the coding primers, one of the tasks to test my coding skills was to use the existing dataset, “babynames,” which tracks the number of babies born with a specific name for each year, and to generate a plot of the number of girls named Amanda from 1880 to the present. Here is the graph that I created to show the popularity of my name over time:
Then, I loaded in spreadsheets from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and tested out my data science coding skills by grouping variables. ADNI is a public database that is a global initiative where researchers are collecting data on AD and sharing it in a group effort to understand and develop a treatment for AD. I was able to organize variables into data tables and graphs in order to interpret the ADNI data more easily. Here is a screenshot of Seraphina and I looking at a data table that I created that filtered out males and females in comparison to other variables, such as their diagnosis, their age, their marital status, and more:
Two of the most interesting variables that we analyzed were the levels of amyloid and levels of tau in the cerebrospinal fluid that was collected for an ADNI study. Looking specifically at the amyloid levels, people with AD often have more amyloid plaque in their brain than people who are cognitively normal (CN). Cerebrospinal fluid surrounds, cushions, and protects the brain and spinal cord, and when high levels of amyloid are found in the brain, low levels of amyloid are found in the cerebrospinal fluid. I made two boxplot graphs to compare the levels of amyloid and the levels of tau in people with a presumed AD diagnosis to people who were CN. Here is the boxplot comparing the amyloid levels in the cerebrospinal fluid of people who have presumed AD vs people who are CN:
Additionally, I was able to analyze the findings of my literature search using the coding that I learned. After uploading my spreadsheet into RStudio, I chose specific diffusion MRI software and tools and created a graph where Seraphina and I were able to see which software and tools had upward trends in popularity. I learned that the software FSL and its tools such as the DTI Toolbox, FMRIB’s Diffusion Toolbox (FDT), the Brain Extraction Tool (BET), and more was one of the most popular softwares used for diffusion MRI scanning in the past decade. Also, MRtrix was another popular software that I was able to utilize in my own quality assessment of diffusion MRI scans.
My third project using RStudio was to determine which diffusion MRI scans were the best to use for my quality assessment. This decision was based on the number of gradient directions and the number of scans taken within one manufacturer. After collecting that information, we used the software MRtrix to view 57 diffusion MRI and anatomical MRI scans from Siemens with 126 gradient directions taken on a MRI scanner with a field strength of 3.0 Tesla in MRview.
In my quality assessment, I looked at the sagittal, coronal, and axial views of the brain. For the diffusion MRI scans, I took note of any volumes where the brain moved and where signal dropout occurred and how severe the signal dropout was. For the anatomical MRI scans, I looked for scans with clean edges and without any smearing, shaking, or blurring.
Conducting the quality assessment was difficult at first because the ADNI scans hadn’t been corrected for distortion, and in a lot of the scans, the brain was pulsing, which made it challenging to distinguish between if the brain was pulsing or distorted or if the subject was moving their head. In our meetings, Seraphina and I went over scans that I had questions about and identified common artifacts that were occuring in the data set. After processing the scans, Seraphina will be able to use the high quality scans in her dissertation. Here is an example of an anatomical MRI scan:
Example of high quality diffusion MRI scan
Example of low quality diffusion MRI scan with signal dropout
Here is a sample of my diffusion MRI quality assessment:
Seraphina has taught me so much about the brain, AD, diffusion MRI, and the field of neuroscience. I also learned many valuable skills that I can use in the future for conducting research, such as how to efficiently read and write a research paper, complete a literature search, code for data science analytics, and execute a quality assessment on diffusion MRI scans. I want to thank Seraphina and Mr. Schlenker for providing me with this amazing opportunity. After this internship, one of my new, future goals is to conduct my own research and to become published. Seraphina’s work has inspired me to further explore my passion for neuroscience and research.
