Caroline Casey - Week 4 - Complex Systems Group: Brain Surface Mapping, PowerPoint, and Writing a Paper - University of Pennsylvania

This week was very exciting and eventful! First, I think I should mention that I am actually completing 10 weeks in the lab, I am excited to be spending so much time researching! This week, Dr. Bassett told me to start writing down everything I have done in the form of a paper. I think I may be getting published since Dr. Bassett wants me to write a paper and since I am now analyzing the results of my project!

On Monday, I re-did from the beginning all the computations I completed in the past few weeks. I am analyzing 2 sequences for each of the subjects and when I first completed the t-test of the movement time data to find the interference values, Dr. Bassett and I decided to average the t-values of the two sequences together. However, Dr. Bassett and I decided to go back to the beginning and separate the 2 sequences, and keep them separate throughout all the computations. We did this because researchers and other scientists would be very critical of the fact that I averaged the statistics. Statistic analysis is always supposed to be the last step. I completed doing all of those computations in a day as opposed to a couple weeks because I became so familiar with the steps I needed to take as well as the algorithms needed. However, I realized that while the behavior data (the recorded movement time) was separated into 2 sequences, the brain data (the adjacency matrices of the fMRI scans) were not separated into 2 sequences. Therefore, we were forced to combine the movement time of the 2 sequences together using another method (which I completed on Tuesday).

On Monday afternoon, we had a lab meeting accompanied by a lecture by a mathematician and topologist at Penn, Rob Ghrist. Rob talked about his work as a topologist. He also gave us an overview of his book that will be coming out in the fall. It was a very interesting lecture, his work is amazing and it was great to hear more about a field that I am not very familiar with!

Tuesday, Dr. Bassett and I decided that the best way to combine the movement time data of sequence 1 and sequence 2 would be to first complete a correlation between the two sequences for the 4 different scenarios in order to determine whether or not we can combine the 2 sequences. Just to remind you, movement time is the time taken to complete the sequence. The movement time data (the 4 scenarios) we have for each subject is: pre-scanning session 2, post-scanning session 2, pre-scanning session 3, and post-scanning session 3, for both sequences. It is important to note that there were many trials of each sequence for each of the scenarios. The subjects came in for a scanning session every two weeks, where the brain data (fMRI) was collected. The interesting point is that the interference effect exists even after the scanning session. The interference effect exists when the post movement time is higher than the pre movement time. So, I completed a Pearson's correlation between sequence 1 and sequence 2 for the 4 scenarios listed above. There was a significant positive correlation, which meant that I could combine the two sequences together.

The next step involved creating 4 long vectors with the movement times of the various trials for both sequence 1 and sequence 2 combined, for each of the 20 subjects. For each subject, one vector held the pre-scanning session 2 data, another vector held the post-scanning session 2 data, the third vector held the pre-scanning session 3 data, and the fourth vector held the post-scanning session 3 data. Then, a t-test was performed between the pre-session 2 vector and the post-session 2 vector, and another t-test between the pre-session 3 vector and the post-session 3 vector for each of the subjects. The t-values for each subject were then placed into 2 interference-value vectors (1x20), one with the session 2 t-values and the other with the session 3 t-values. I then went on to complete the exact same steps I did twice previously. The next step involved computing a Pearson's correlation between the interference vectors and the brain data for 4 different scenarios: interference values session 2 with scan 1, interference values session 2 with scan 2, interference values session 3 with scan 2, and interference values session 3 with scan 3. I then found the significant edges as I did before and finally visualized the data as I did before. While it sounds like it would only take a couple hours to complete, it took Tuesday and most of Wednesday to complete all of those steps.

On Wednesday afternoon, I started analyzing the now correct networks I have. I used a community detection algorithm, however, the community structure is very weak in the networks. The next analysis technique I completed was finding the degree of each of the nodes in the networks. The degree of the nodes is simply how many other nodes it is connected to in the network. I created 4 degree vectors for each of the 4 correlation scenarios I computed earlier in order to find the edges: interference values session 2 with scan 1, interference values session 2 with scan 2, interference values session 3 with scan 2, and interference values session 3 with scan 3. This makes it easier to see the connections within the network. I continued working on visualizing the networks and looking at the degree distribution of the nodes on Thursday.

I have had some results in finding areas in the brain that have the same degree throughout all 4 scenarios. I am trying to identify whether or not the interference effect is due to anxiety, environment, or different biomechanics. I can identify this by seeing what regions of the brain seem to play a critical role in the interference effect and figuring out whether or not those regions are responsible for any feelings of stress or anxiety. If no regions seem to fit an anxiety or stress function, then we can conclude that the interference effect is due to different biomechanics.

Thursday, I also computed the skew of the 4 degree vectors I created and also found the p-values for the skews. The histograms of the degree vectors were significant and highly skewed. On Thursday, Dr. Bassett said that since I already have results, I should create a PowerPoint of the motivation behind the research I was conducting as well as everything I have done so far in completing my project. I will have a Skype conversation with the researcher who collected the brain data to share the PowerPoint and go over my results. I am really looking forward to sharing what I have discovered so far!

On Friday, one of the Graduate students, Muzhi showed me how to use a brain surface mapping program that maps the nodes onto an actual brain structure. I began looking at those images to see if there are any trends in the images. I also worked all day on the PowerPoint and completed it, all 37 slides! I also worked on the paper that I am writing about my research. I have finished the Introduction and Methods sections and started working on the Discussion section.

This is the brain surface mapping I completed of one of the scenarios. The nodes are colored based on the degree (see gradient image below the next two images).  This is the axial view.

This is the brain surface mapping I completed of one of the scenarios. The nodes are colored based on the degree (see gradient image below the next image).  This is the coronal view.

This is the brain surface mapping I completed of one of the scenarios. The nodes are colored based on the degree (see gradient image below).  This is the saggital view.

This is the degree gradient of the scenario that is shown above in the brain images. The numbers indicate the degree number, the color represents the color of the node with the given degree number.

I am so excited to see the results of my research and to start sharing my discoveries! I am loving the experience!