Tuesday:
Today we had our first fMRI fundamentals talk. Dr. Hirsch taught it, and it was really interesting. We learned about some brain theories and how the fMRI works. I am going to try and explain the whole lecture in another post.
After the talk, about half of the summer students and I went back to the area where I normally sit. (there are two areas in the lab, the main area and the back area. I sit in the main area because that's where Spiro is, but there is more room in the back so a lot of summer students work there.) I started working more on choosing my stimuli, which, at a certain point, became a group activity. After a while, we all went back to doing our own things. By the end of the day, I narrowed my stimuli down to about 80 pieces and started to shorten them to 10-15 second long clips.
Choosing the stimuli was tough. As I mentioned in a previous entry, I cannot assign emotions to the pieces that I use: they must be chosen by a population similar to my subjects. However, in choosing the pieces to put into the pre-study, I have to choose pieces with relatively obvious and clear emotions. Out of the 110 pieces I started with, I felt that all of them had an easily defined emotion. I did my first filter through by deciding that each piece needed to have at contain at least one minute defined by a single emotion. This got rid of easily 20 songs.
Friday, July 9, 2010
Thursday, July 1, 2010
Monday
Now back to a semi-daily log:
Monday:
Today was a crazy day. Slight backstory, on Thursday I went to talk with dr. Hirsch about my study, but she was super busy. She said she we could talk friday, but I wasn't coming in on friday, so we planned to talk today. Almost all day her doors were closed. There was another summer student waiting to talk to her about his study, which is about the perception of pain as a person giving pain and as an observer. It was interesting and, luckily, only used a simulation of pain infliction using pictures. Had the pain not been simulated, the experiment would basically be the famous Milgram experiment, just with 2 people doing the shocking instead of 1. We talked a lot about his experiment, my experiment, and other stuff. Eventually dr. Hirsch came outside, told the other student that she was super busy but could meet with him for a few minutes and said nothing to me. At this point, I decided to go talk to the grad student in charge of summer students, Grace, about my study. When I went to talk to her, she basically told me that, because I am not required to do a study for a specific program, I can't run one. So for about an hour, I had nothing to do. Spiro wasn't there, so I couldn't get any new work from him, and I didn't want to work on my study if I wasn't going to be able to continue with it. Eventually, Grace came back and told me that I could, in fact, work on my study, but the students who are required to do a study get priority for things over me.
For the rest of the day, I did reading and stimulus-finding. (This was actually the day I changed from decision making to math.) What's really nice about the lab is that basically everyone is happy to help with whatever you want. I asked Spiro a ton of experimental design questions, and not only did he not mind me asking, he also knew the answers. Additionally, I was able to bounce a lot of my ideas off of other summer students; my final design was helped along a lot by Evelyn.
Based on today's developments, I think I will be a bit behind schedule for finishing running my pre-study by the end of this week, but I feel like it's not actually that big of a deal.
Monday:
Today was a crazy day. Slight backstory, on Thursday I went to talk with dr. Hirsch about my study, but she was super busy. She said she we could talk friday, but I wasn't coming in on friday, so we planned to talk today. Almost all day her doors were closed. There was another summer student waiting to talk to her about his study, which is about the perception of pain as a person giving pain and as an observer. It was interesting and, luckily, only used a simulation of pain infliction using pictures. Had the pain not been simulated, the experiment would basically be the famous Milgram experiment, just with 2 people doing the shocking instead of 1. We talked a lot about his experiment, my experiment, and other stuff. Eventually dr. Hirsch came outside, told the other student that she was super busy but could meet with him for a few minutes and said nothing to me. At this point, I decided to go talk to the grad student in charge of summer students, Grace, about my study. When I went to talk to her, she basically told me that, because I am not required to do a study for a specific program, I can't run one. So for about an hour, I had nothing to do. Spiro wasn't there, so I couldn't get any new work from him, and I didn't want to work on my study if I wasn't going to be able to continue with it. Eventually, Grace came back and told me that I could, in fact, work on my study, but the students who are required to do a study get priority for things over me.
For the rest of the day, I did reading and stimulus-finding. (This was actually the day I changed from decision making to math.) What's really nice about the lab is that basically everyone is happy to help with whatever you want. I asked Spiro a ton of experimental design questions, and not only did he not mind me asking, he also knew the answers. Additionally, I was able to bounce a lot of my ideas off of other summer students; my final design was helped along a lot by Evelyn.
Based on today's developments, I think I will be a bit behind schedule for finishing running my pre-study by the end of this week, but I feel like it's not actually that big of a deal.
Weeks 4 & 5
Weeks 4 & 5: Research
When I got to the lab on Monday, I started thinking about what I was going to work on for the rest of the summer. I really liked all the programming stuff I was doing, but I wanted to also do some research, and not just work on the results of Spiro's work. So I decided I wanted to run my own study, the problem was, I had no idea what I wanted to study! My timing plan for the week was to do general research for a few days and find a topic by Tuesday or Wednesday, and then have a project plan by Friday. I went through a RIDICULOUS amount of study ideas; everything from using music and American Sign Language to help autistic children to communicate, to researching handedness, to researching the effects of amount of music on advertising. By tuesday (surprisingly) I decided I was going to research the effect of musical emotion (the perceived emotion of a given piece) on decision making. (Quick side note: my ideal research project was to have something that I could do a behavioral study on this summer and look at other results in the lab so that I could go up to school in september, present my results to a lab, and then do the imaging part of the study at UR. This idea fit all those criteria.) The advertising idea was the precursor to the decision making idea. For the rest of the week, I worked on study planning.
The two biggest parts of this study are obviously the music and the decision, so right away I started compiling a playlist of emotional music. In most studies that deal with emotion, the 4 goto emotions are happy, sad, angry, and fearful. However, I had an idea that, for a decision making task, a more tense piece of music may rush the decision while a more relaxed piece would make the subject think more. So, instead of angry and fearful, I decided to use tense and peaceful.
In papers I have read about music emotion studies, the general consensus is that the emotion of the stimuli cannot be judged solely by the researcher, rather it must be judged by a population similar to the subjects the will be used in the experiment. This means that the first thing I have to do is run a pre-study with a whole bunch of musical stimuli.
While I prep and run the pre-study, I will plan the actual study. The decision to be made in the experiment was making everything difficult. I wanted a decision (could be high or low risk, doesn't matter) that had two option: the first looks like the better choice on the surface, but when you think about it, it is definitely the worse of the two. The other choice is the opposite: looks bad on the surface but after some thought, is actually pretty good. It is VERY difficult to find a decision like this. In fact, I found it nearly impossible. After a few days of research, I gave up on the decision and instead focused on task performance; specifically math. The experiment now studies the effect of emotional arousal level on simple math performance. Earlier in the summer, I read a thesis that discussed the effects of musical arousal (in terms of tempo, volume, and likeablility) on task performance; specifically math and memory. I had a lot of issues with this study, but that doesn't really matter. What's important is that I am using her idea of musical arousal, but instead of defining it by volume or tempo, I am defining it by emotion (which include a whole new brain area).
I am running the study in a simple block design that is streamlined for good fMRI performance. The way the block design works is that I have music emotion blocks, problems set blocks, and rest blocks. Music emotion blocks are simply a piece of music with a given emotion that is about 1 minute long. Problem set blocks are just a set of arithmetic problems. The difficulty will be predetermined by a computer test I am writing. Each p-set will have about 150 problems in it (I think) because the task is to get as many problems as is possible done during the allotted time. A rest block is about 20 seconds long, and is just silence with a blank screen. Each trial is set up like so: a music emotion block and a p-set block are paired and run simultaneously (the piece of music plays while the subject answers problems). After the minute is up, there is a rest block. This is repeated 9 more times (for different conditions). Finally, the entire experiment is repeated to test for accuracy (that may not actually happen).
This is streamlined for fMRI because it has constant stimulus/rest blocks. In the scanner, a study must be designed in this way in order to actually see the activation. By inserting a rest period between each trial, it allows the brain to return to baseline (how the brain fires when nothing is happening). However, before I can run any of the experiment, I still need to run the pre-study.
When I got to the lab on Monday, I started thinking about what I was going to work on for the rest of the summer. I really liked all the programming stuff I was doing, but I wanted to also do some research, and not just work on the results of Spiro's work. So I decided I wanted to run my own study, the problem was, I had no idea what I wanted to study! My timing plan for the week was to do general research for a few days and find a topic by Tuesday or Wednesday, and then have a project plan by Friday. I went through a RIDICULOUS amount of study ideas; everything from using music and American Sign Language to help autistic children to communicate, to researching handedness, to researching the effects of amount of music on advertising. By tuesday (surprisingly) I decided I was going to research the effect of musical emotion (the perceived emotion of a given piece) on decision making. (Quick side note: my ideal research project was to have something that I could do a behavioral study on this summer and look at other results in the lab so that I could go up to school in september, present my results to a lab, and then do the imaging part of the study at UR. This idea fit all those criteria.) The advertising idea was the precursor to the decision making idea. For the rest of the week, I worked on study planning.
The two biggest parts of this study are obviously the music and the decision, so right away I started compiling a playlist of emotional music. In most studies that deal with emotion, the 4 goto emotions are happy, sad, angry, and fearful. However, I had an idea that, for a decision making task, a more tense piece of music may rush the decision while a more relaxed piece would make the subject think more. So, instead of angry and fearful, I decided to use tense and peaceful.
In papers I have read about music emotion studies, the general consensus is that the emotion of the stimuli cannot be judged solely by the researcher, rather it must be judged by a population similar to the subjects the will be used in the experiment. This means that the first thing I have to do is run a pre-study with a whole bunch of musical stimuli.
While I prep and run the pre-study, I will plan the actual study. The decision to be made in the experiment was making everything difficult. I wanted a decision (could be high or low risk, doesn't matter) that had two option: the first looks like the better choice on the surface, but when you think about it, it is definitely the worse of the two. The other choice is the opposite: looks bad on the surface but after some thought, is actually pretty good. It is VERY difficult to find a decision like this. In fact, I found it nearly impossible. After a few days of research, I gave up on the decision and instead focused on task performance; specifically math. The experiment now studies the effect of emotional arousal level on simple math performance. Earlier in the summer, I read a thesis that discussed the effects of musical arousal (in terms of tempo, volume, and likeablility) on task performance; specifically math and memory. I had a lot of issues with this study, but that doesn't really matter. What's important is that I am using her idea of musical arousal, but instead of defining it by volume or tempo, I am defining it by emotion (which include a whole new brain area).
I am running the study in a simple block design that is streamlined for good fMRI performance. The way the block design works is that I have music emotion blocks, problems set blocks, and rest blocks. Music emotion blocks are simply a piece of music with a given emotion that is about 1 minute long. Problem set blocks are just a set of arithmetic problems. The difficulty will be predetermined by a computer test I am writing. Each p-set will have about 150 problems in it (I think) because the task is to get as many problems as is possible done during the allotted time. A rest block is about 20 seconds long, and is just silence with a blank screen. Each trial is set up like so: a music emotion block and a p-set block are paired and run simultaneously (the piece of music plays while the subject answers problems). After the minute is up, there is a rest block. This is repeated 9 more times (for different conditions). Finally, the entire experiment is repeated to test for accuracy (that may not actually happen).
This is streamlined for fMRI because it has constant stimulus/rest blocks. In the scanner, a study must be designed in this way in order to actually see the activation. By inserting a rest period between each trial, it allows the brain to return to baseline (how the brain fires when nothing is happening). However, before I can run any of the experiment, I still need to run the pre-study.
Week 3 in Review
I know I haven't posted an entry in an extremely long time, so I am going to write a couple of long entries that sum up what I've been doing at the lab.
Week 3: Scripting
Last I posted, I explained what spiros script does, the work he is doing, and what I did on the script. During the rest of the week, I worked a lot more on scripting. There was a small problem with the script where any voxel with a value of zero (which was most of them) turned gray and made the image extremely hard to read. It was a simple fix, but it took forever to figure out!
On FSL there is an awesome tool called "atlases" (I may have already talked about them?) With an atlas, you can take any normalized brain image, click on any spot on the image, and the atlas will tell you what part of the brain you are looking at. While working on the node script, I thought it would be a really cool feature if there was an atlas for the nodes, so I started to look into how to make an atlas. Unfortunately, there are no online resources for creating atlases in fsl, and the users manual for the atlas tool has a note on the bottom that says: HINTS AND TIPS FOR ATLAS DEVELOPERS. And then the page ends. So, because I had not much better to do, I found where the atlas info was stored on the computer, and tried to figure out how to write one. A basic atlas consists of two files: a reference image and a data file. The reference image is a standard fsl file(.nii.gz) but the data file is written in XML, which I don't know and have basically no use for. So instead of learning xml, I just copied and pasted the contents of one file and filled in the blanks for my atlas. The way that the file works, is that on each line there are the x, y, and z coordinates of the brain area and then a title. The file then finds the coordinates on the reference image and determines how large the brain area is based on intensity values. Once I figured all this out, I was under the impression that I would have to manually open each of the 369 node images and define the center of each node for the data file. I tried that for about 8 images, and then decided that there must be an easier way. I came in the next day, and pieced together a few scripts that could find image centers to make an "atlas creator" script. I paired the atlas creator with the node script, and 20 minutes later (running the script took FOREVER) I had a personalized atlas that showed all of the nodes! The last step was to create an installer package that would create the necessary files and move them to their correct locations.
The next thing Spiro asked me to do was to see if I could somehow show the connection on each node, or make a 3D representation of the nodes and connections. I tried, but didn't get very far. One of the programs involved learning the C programming language, which I did over the weekend. However, when I got back to the lab on Monday, I decided I wanted to design my own study.
Week 3: Scripting
Last I posted, I explained what spiros script does, the work he is doing, and what I did on the script. During the rest of the week, I worked a lot more on scripting. There was a small problem with the script where any voxel with a value of zero (which was most of them) turned gray and made the image extremely hard to read. It was a simple fix, but it took forever to figure out!
On FSL there is an awesome tool called "atlases" (I may have already talked about them?) With an atlas, you can take any normalized brain image, click on any spot on the image, and the atlas will tell you what part of the brain you are looking at. While working on the node script, I thought it would be a really cool feature if there was an atlas for the nodes, so I started to look into how to make an atlas. Unfortunately, there are no online resources for creating atlases in fsl, and the users manual for the atlas tool has a note on the bottom that says: HINTS AND TIPS FOR ATLAS DEVELOPERS. And then the page ends. So, because I had not much better to do, I found where the atlas info was stored on the computer, and tried to figure out how to write one. A basic atlas consists of two files: a reference image and a data file. The reference image is a standard fsl file(.nii.gz) but the data file is written in XML, which I don't know and have basically no use for. So instead of learning xml, I just copied and pasted the contents of one file and filled in the blanks for my atlas. The way that the file works, is that on each line there are the x, y, and z coordinates of the brain area and then a title. The file then finds the coordinates on the reference image and determines how large the brain area is based on intensity values. Once I figured all this out, I was under the impression that I would have to manually open each of the 369 node images and define the center of each node for the data file. I tried that for about 8 images, and then decided that there must be an easier way. I came in the next day, and pieced together a few scripts that could find image centers to make an "atlas creator" script. I paired the atlas creator with the node script, and 20 minutes later (running the script took FOREVER) I had a personalized atlas that showed all of the nodes! The last step was to create an installer package that would create the necessary files and move them to their correct locations.
The next thing Spiro asked me to do was to see if I could somehow show the connection on each node, or make a 3D representation of the nodes and connections. I tried, but didn't get very far. One of the programs involved learning the C programming language, which I did over the weekend. However, when I got back to the lab on Monday, I decided I wanted to design my own study.
Monday, June 14, 2010
Day 8
It's been a while since I've posted. I wasn't at the lab on Thursday and Friday, but Spiro gave me the script to work on over the weekend. I'm going to try and explain what this script does and what I did today, but if it doesn't make sense please leave a comment or something so I can clear it up. I believe I have to present it to some people on Thursday.
I'll start by explaining Spiro's work. Spiro is researching graph theory in neuroscience. Basically, the whole brain is connected in two ways: physical connections and functional connections. Using DTI (mentioned in previous posts) we can easily map physical connections, and using fMRI we can map functional connections. Spiro is using both DTI and fMRI to map the connections between certain areas of the brain during fear responses. (Side note: Fear is very commonly studied in neuroscience labs. The brain and body react to fear in many quantifiable ways, so it is nice to use for mapping certain connections in the brain. Often, studies are done with people with and without high anxiety or anxiety disorders, because they have different pathways for fear.) In his research, Spiro has acquired data on how many connections certain "nodes" of the brain have at both baseline and during fear responses. (A node is part of a large structure of the brain. There are normally four nodes in each known structure.) The point of this script is to put all the data from these nodes and put them onto the same brain. Each node is recorded on an image. That image has values (called "Intensity Values") at every point (voxel) on the brain, however, not all of the points matter. The first thing that this script does, is set every voxel below a certain intensity (defined in the script) to zero. Next, it sets every value above zero to be one. These two steps create an image that highlights only the node, with one value in each highlighted voxel. The script then multiplies every value in the image by the number of connections (defined by a vector the user inputs). The script repeats these steps for all 369 nodes, adding each one to a final image. Finally, that image is overlaid onto a standardized brain. (Did that make sense?)
Today, I took the script, which initially took 4 minutes, 40 seconds to run, and sped it up to take only 2:18. I did that by combining a few steps that created temporary images for each step. By not creating and deleting images on each step, the script runs much faster. I also made it so that the script automatically opens up the image (color-coded) once it finishes running. Finally, I added in the ability to choose your own threshold value (the number below which all values turn to zero). It took me easily a half hour to figure out how to input the threshold value. The script is now done, which is awesome!
Today we also saw one of the post docs present her research. Her presentation was about an hour long, and was a bout fear connections in mice. It was an extremely interesting talk. She was doing a practice talk for her job interview, which is tomorrow.
Tomorrow I may begin to figure out what I am going to work on for the rest of the summer. Woo hoo!
I'll start by explaining Spiro's work. Spiro is researching graph theory in neuroscience. Basically, the whole brain is connected in two ways: physical connections and functional connections. Using DTI (mentioned in previous posts) we can easily map physical connections, and using fMRI we can map functional connections. Spiro is using both DTI and fMRI to map the connections between certain areas of the brain during fear responses. (Side note: Fear is very commonly studied in neuroscience labs. The brain and body react to fear in many quantifiable ways, so it is nice to use for mapping certain connections in the brain. Often, studies are done with people with and without high anxiety or anxiety disorders, because they have different pathways for fear.) In his research, Spiro has acquired data on how many connections certain "nodes" of the brain have at both baseline and during fear responses. (A node is part of a large structure of the brain. There are normally four nodes in each known structure.) The point of this script is to put all the data from these nodes and put them onto the same brain. Each node is recorded on an image. That image has values (called "Intensity Values") at every point (voxel) on the brain, however, not all of the points matter. The first thing that this script does, is set every voxel below a certain intensity (defined in the script) to zero. Next, it sets every value above zero to be one. These two steps create an image that highlights only the node, with one value in each highlighted voxel. The script then multiplies every value in the image by the number of connections (defined by a vector the user inputs). The script repeats these steps for all 369 nodes, adding each one to a final image. Finally, that image is overlaid onto a standardized brain. (Did that make sense?)
Today, I took the script, which initially took 4 minutes, 40 seconds to run, and sped it up to take only 2:18. I did that by combining a few steps that created temporary images for each step. By not creating and deleting images on each step, the script runs much faster. I also made it so that the script automatically opens up the image (color-coded) once it finishes running. Finally, I added in the ability to choose your own threshold value (the number below which all values turn to zero). It took me easily a half hour to figure out how to input the threshold value. The script is now done, which is awesome!
Today we also saw one of the post docs present her research. Her presentation was about an hour long, and was a bout fear connections in mice. It was an extremely interesting talk. She was doing a practice talk for her job interview, which is tomorrow.
Tomorrow I may begin to figure out what I am going to work on for the rest of the summer. Woo hoo!
Wednesday, June 9, 2010
Day 7
Today was a relatively boring day. I spent almost all of my day learning FSL, then I started to learn bash scripting. When Spiro came in, he told me that he started to work on the script we are writing and showed me what he had done so far. I actually understood most of what the script was doing and how it was doing it, which was new! Spiro sent me the script/data so that I could improve on it. And that was about all I did today.
The one cool thing that happened today was that I got pictures of my brain from the scan I did last week! I posted them here.
3D Picture 2
3D Picture 3
3D Picture 4
The one cool thing that happened today was that I got pictures of my brain from the scan I did last week! I posted them here.
3D Picture 2
3D Picture 3
3D Picture 4
Day 6
When I arrived at the lab today, I walked into the area where I normally sit and saw three people I didn't know sitting there. At first I assumed they were all summer students; only one of them was. The other two were a researcher and subject involved in a very cool study involving DDR (yes, Dance Dance Revolution). The study is researching how DDR can be used as effective therapy in Parkinsons patients. The theory is that DDR has about the same amount of physical exertion as standard Treadmill Training does. Treadmill Training involves having the patient walk on a treadmill while their weight is supported by a harness. Eventually, they begin to support more and more of their own weight and increase their speed. The problem with TT is that it is boring and it is difficult to see immediate improvements. On the surface, DDR training solves both of these issues: it is engaging, and after each song you are given a score that tells you exactly how you did. Additionally, it is theorized that DDR will actually work better than TT because it involves additional coding. The example I was given today, was this: in the brain, different areas are all connected. Say, for example, area A is connected to B, and B is connected to C. As a result, A is connected to C through B. In Parkinsons patients, B is broken, so A and C are no longer connected. The brain must find an alternate pathway to connect the two areas (often motor areas). TT does this by forcing the legs to work, which then forces the brain to make new connections to compensate (often the therapist will have to physically move the patient's legs themselves). This often takes a long time because very few areas of the brain are working together. With DDR, the motor functions (pushing the arrows on the pad) are accompanied by both visual stimuli (arrows on the screen) and audio stimuli (the music). In this way, the motor system, visual system, auditory system, as well as others are all working together to forge new pathways in the brain.
At this point in the study, they are working on getting their control data. Additionally, they are using the control subjects to prepare for using actual Parkinsons patients later this summer. They have two control groups at the moment: healthy young subjects and healthy elderly subjects. I believe the reason for this is to get a true set of control data (young and healthy) and then get a more similar control group to the experimental group (elderly and healthy).
The subject that was being run today was a young, healthy woman, back for her second round of scanning. Each subject is scanned multiple times. First, they are scanned before any DDR training is done. This gives each person a baseline scan. The scan has 11 parts. First, the subject gets a basic structural scan (this only happens in the first set of scans). Next, the subject gets a basic DTI scan. Then the DDR tasks start. While in the scanner, the subject has a foot pad that has two buttons. In the next group of scans, the subject plays DDR using these two buttons. In scan three, the subject plays a normal round of DDR, except for the fact that when round symbols come up on the screen, they do not press the buttons. In scan four, the subject must press the buttons in time with the music (left, right, left, right). In scans five and six are exactly the same as three and four, just without the music. The next five scans are exactly the same as scans 2-6. They do this so they can be sure they have accurate data. After a month or so (and approximately 35+ hours of DDR training) the subjects come back to the lab and do that same set of scans again. I believe there is only one follow-up scan. The whole experiment takes forever!
I very much like the concept of this study, however I did have one issue: in the trials with no music, the factor of subjects imagining the music in their head is ignored. I feel that this is a HUGE confound that needs to be taken into account in some way. Adam (the guy running the study) said that they had been trying to think of a way to account for it, but had not figured one out yet.
After watching the first run of the scans (scans 1-5) I went back to get some of my own work done, and once I helped the other summer students with various computer issues, I was able to do that. FSL (the program I am learning) is REALLY cool. It is really nice to finally see some real brains! I started off doing some viewing of images and other basic maneuvers in the program. Some of the notation in the software is really confusing. For example, every scan, no matter what type, has an Intensity value for each voxel. Depending on the type of scan, this Intensity value means something completely different, so you just have to know what it means in each scan you are looking at. It's pretty annoying. The coolest thing I found today, though, was the Atlas tool. The atlas tool allows you to point to any area in a normalized brain (a template brain) and it will tell you what that part of the brain is (ie. thalamus, amygdala, putamen, etc.). You can even set it up to highlight all of the areas in whatever atlases you choose. For the rest of my time there today I just turned on the highlighting for different atlases and learned where different parts of the brain are, how big they are, etc. It was really cool! Tomorrow I need to make a lot more of a dent is the FSL tutorial so that I can start working on the script with Spiro.
At this point in the study, they are working on getting their control data. Additionally, they are using the control subjects to prepare for using actual Parkinsons patients later this summer. They have two control groups at the moment: healthy young subjects and healthy elderly subjects. I believe the reason for this is to get a true set of control data (young and healthy) and then get a more similar control group to the experimental group (elderly and healthy).
The subject that was being run today was a young, healthy woman, back for her second round of scanning. Each subject is scanned multiple times. First, they are scanned before any DDR training is done. This gives each person a baseline scan. The scan has 11 parts. First, the subject gets a basic structural scan (this only happens in the first set of scans). Next, the subject gets a basic DTI scan. Then the DDR tasks start. While in the scanner, the subject has a foot pad that has two buttons. In the next group of scans, the subject plays DDR using these two buttons. In scan three, the subject plays a normal round of DDR, except for the fact that when round symbols come up on the screen, they do not press the buttons. In scan four, the subject must press the buttons in time with the music (left, right, left, right). In scans five and six are exactly the same as three and four, just without the music. The next five scans are exactly the same as scans 2-6. They do this so they can be sure they have accurate data. After a month or so (and approximately 35+ hours of DDR training) the subjects come back to the lab and do that same set of scans again. I believe there is only one follow-up scan. The whole experiment takes forever!
I very much like the concept of this study, however I did have one issue: in the trials with no music, the factor of subjects imagining the music in their head is ignored. I feel that this is a HUGE confound that needs to be taken into account in some way. Adam (the guy running the study) said that they had been trying to think of a way to account for it, but had not figured one out yet.
After watching the first run of the scans (scans 1-5) I went back to get some of my own work done, and once I helped the other summer students with various computer issues, I was able to do that. FSL (the program I am learning) is REALLY cool. It is really nice to finally see some real brains! I started off doing some viewing of images and other basic maneuvers in the program. Some of the notation in the software is really confusing. For example, every scan, no matter what type, has an Intensity value for each voxel. Depending on the type of scan, this Intensity value means something completely different, so you just have to know what it means in each scan you are looking at. It's pretty annoying. The coolest thing I found today, though, was the Atlas tool. The atlas tool allows you to point to any area in a normalized brain (a template brain) and it will tell you what that part of the brain is (ie. thalamus, amygdala, putamen, etc.). You can even set it up to highlight all of the areas in whatever atlases you choose. For the rest of my time there today I just turned on the highlighting for different atlases and learned where different parts of the brain are, how big they are, etc. It was really cool! Tomorrow I need to make a lot more of a dent is the FSL tutorial so that I can start working on the script with Spiro.
Monday, June 7, 2010
Day 5
Today starts week two of my internship. I'm going to start this entry with an introduction, because I never did that earlier. Also, if anything I say in this blog is unclear, confusing, etc. please tell me in the comments so that I can clear it up.
This summer I am interning at an fMRI lab at Columbia. I will be working with grad students in the lab and contributing to current research. It's pretty cool. You may be saying right now, "Well, I've heard of an MRI, but how is an fMRI different?" An MRI is the most basic form of magnetic imaging. Standard MRIs (also known as T1 scans) take pictures of slices of the brain. They are mainly used for structural imaging. fMRIs (also known as T2* scans) measure blood flow in the brain. (The 'f' stands for functional) The way that fMRIs work is that they track the BOLD Signal (Blood Oxygen Level Dependency) in each voxel (3D pixel) of the brain over time. In other words, it tracks where blood goes when you do a certain task. In this way, we can see what specific parts of the brain are used for. fMRIs are often used for research, but they are also commonly used pre-brain surgery to see what parts of the patients brain are used for what. For example, most people process language on the left side of the brain, but if the patient processes language on the right side, the person operating needs to know that. (Fun fact for those who know what WADA is, the WADA test is still the gold standard for pre-surgery testing. Even though Pouget said it is rarely used, he was wrong.)
One of the tricker things about fMRI is the hemodynamic lag, or the time it takes for blood to go to an activated brain area. Once a part of the brain is activated by a certain stimulus, it takes about 1-5 seconds for blood flow in that area to increase. This must be factored in and averaged out while analyzing the results. I will talk more about this later when I am actually looking at real fMRI data.
Today the rest of the summer students came. I like them! After Grace (a grad student in the lab who is head of summer students) talked to us for a bit, I started to teach the three other students how to use UNIX and some MATLAB commands. I was impressed by how well I knew the material. After a while, Grace told us that there was a study subject who was about to go into the scanner and, if we wanted to, we could go watch the study.
This study had to do with how exposure to emotional faces can affect visual perception, specifically in the periphery. The task was extremely difficult (in my opinion). First, a face which portrayed an emotion (happy, fearful, or neutral) flashed on the screen for less than a second (maybe even less than half a second.) Immediately after, eight groups of lines showed up on the screen for about the same amount of time. Half of the lines were red and the other half were blue; they were arranged in a circle around the center of the screen. One of the groups of blue lines were slanted either left or right, and the task was to determine that direction.
The first section of the experiment was basically a practice round, except that the program calibrated the angle at which the lines were slanted so that the task was not too easy and not too difficult. After that, the subject had to do four trails of about ten minutes each. The task was exactly the same for all four trials. Next, the subject got a retinotopic scan, which mapped out how his visual cortex is set up. Lastly, as in all fMRI scans, the subject got a structural scan. (The reason that everyone gets a structural scan is so that the results can be projected onto a high resolution picture of the subject's brain.) That was the end of the first of two days of scanning for this subject. In addition to the fMRI, the subject's heart rate and galvanic skin response were measured, and he was also connected to an eye tracker. (Given that the study was measuring his periphery, any trial in which his eyes left the middle of the screen will be discarded.)
While we were watching the scan, we were able to ask Michelle (the grad student running this study) a lot of questions about her study, how the stimuli worked, how the scanner worked, etc. She and the tech, Steve, answered every one of our questions in great detail, which was extremely helpful. It is possible that we may be subjects for this study in the future.
I finished the day talking to Spiro about the script we are writing. I haven't gotten to far in the tutorials I need to do, so a lot of what he was saying was way over my head, but I feel like after I run through some of these tutorials tomorrow I will understand it a lot better.
This summer I am interning at an fMRI lab at Columbia. I will be working with grad students in the lab and contributing to current research. It's pretty cool. You may be saying right now, "Well, I've heard of an MRI, but how is an fMRI different?" An MRI is the most basic form of magnetic imaging. Standard MRIs (also known as T1 scans) take pictures of slices of the brain. They are mainly used for structural imaging. fMRIs (also known as T2* scans) measure blood flow in the brain. (The 'f' stands for functional) The way that fMRIs work is that they track the BOLD Signal (Blood Oxygen Level Dependency) in each voxel (3D pixel) of the brain over time. In other words, it tracks where blood goes when you do a certain task. In this way, we can see what specific parts of the brain are used for. fMRIs are often used for research, but they are also commonly used pre-brain surgery to see what parts of the patients brain are used for what. For example, most people process language on the left side of the brain, but if the patient processes language on the right side, the person operating needs to know that. (Fun fact for those who know what WADA is, the WADA test is still the gold standard for pre-surgery testing. Even though Pouget said it is rarely used, he was wrong.)
One of the tricker things about fMRI is the hemodynamic lag, or the time it takes for blood to go to an activated brain area. Once a part of the brain is activated by a certain stimulus, it takes about 1-5 seconds for blood flow in that area to increase. This must be factored in and averaged out while analyzing the results. I will talk more about this later when I am actually looking at real fMRI data.
Today the rest of the summer students came. I like them! After Grace (a grad student in the lab who is head of summer students) talked to us for a bit, I started to teach the three other students how to use UNIX and some MATLAB commands. I was impressed by how well I knew the material. After a while, Grace told us that there was a study subject who was about to go into the scanner and, if we wanted to, we could go watch the study.
This study had to do with how exposure to emotional faces can affect visual perception, specifically in the periphery. The task was extremely difficult (in my opinion). First, a face which portrayed an emotion (happy, fearful, or neutral) flashed on the screen for less than a second (maybe even less than half a second.) Immediately after, eight groups of lines showed up on the screen for about the same amount of time. Half of the lines were red and the other half were blue; they were arranged in a circle around the center of the screen. One of the groups of blue lines were slanted either left or right, and the task was to determine that direction.
The first section of the experiment was basically a practice round, except that the program calibrated the angle at which the lines were slanted so that the task was not too easy and not too difficult. After that, the subject had to do four trails of about ten minutes each. The task was exactly the same for all four trials. Next, the subject got a retinotopic scan, which mapped out how his visual cortex is set up. Lastly, as in all fMRI scans, the subject got a structural scan. (The reason that everyone gets a structural scan is so that the results can be projected onto a high resolution picture of the subject's brain.) That was the end of the first of two days of scanning for this subject. In addition to the fMRI, the subject's heart rate and galvanic skin response were measured, and he was also connected to an eye tracker. (Given that the study was measuring his periphery, any trial in which his eyes left the middle of the screen will be discarded.)
While we were watching the scan, we were able to ask Michelle (the grad student running this study) a lot of questions about her study, how the stimuli worked, how the scanner worked, etc. She and the tech, Steve, answered every one of our questions in great detail, which was extremely helpful. It is possible that we may be subjects for this study in the future.
I finished the day talking to Spiro about the script we are writing. I haven't gotten to far in the tutorials I need to do, so a lot of what he was saying was way over my head, but I feel like after I run through some of these tutorials tomorrow I will understand it a lot better.
Friday, June 4, 2010
Day 4
I woke up this morning to an email from Spiro (the grad student I'm working with) saying that he was most likely coming in to the lab around 3 because the water was being turned off in his apartment from 11 to 2. (He also said there was a possibility that he would wake up early and be in by 11:30, but I knew that would never happen). In his email, he went on to say that, if I was interested, he would tell me about the work that he is doing now (which I had shown some interest in) and discuss some further tasks that could contribute greatly to research in this field. Spiro is working on neurological graph theory (kind of like brain mapping but more). His study is based on seeing what parts of the brain activate in response to fear (stimulus is a face showing the fear emotion) and how those parts are connected. He used people with social anxiety disorders as his experimental group. In his research, he has identified nodes in the brain (I have no idea if they are specific areas or specific neurons. I'll hopefully find out soon.) and has compiled a list of which nodes make connections with other nodes. (That isn't all he has done, but for what I'm doing, that's mostly what is important.) My task in all of this is to help develop a script that can, from what I understand, determine the number of connections per node and then project that information onto the brain using a color spectrum scale. (There is a more scientific name for "color spectrum scale", I just can't remember it now.) At the moment, all I can really do is start learning how to use FSL and SPM (two brain image reading programs) and how to script in MATLAB, but eventually (soon) I will be working with Spiro to create this script. Now, the coolest part about this is that this script is extremely important to graph theory research, so it is possible that if I contribute a good amount to this project, my name may be included in papers relating to it!!!
In regards to the DTI research we started running yesterday, I finally got to see some actual data! We finished running all of the scripts we needed to run on the preliminary data, and I got to see the results. (Quick note, the scripts we are running on this data must be run when all the data is present, which it was not yesterday. We did a practice run of the data using only ten of the 27 subjects, so the results we found weren't really results, just pretty cool brain scans.) The experiment is seeing what parts of the brain differ between highly skilled musicians and non-musicians. In the final images, we saw an averaged version of all the data projected onto one "normalized" brain. (I know it sounds weird, but I swear it works) All of the white matter tracks show up as green. Once the experimental conditions are added, some of the green either turns red or blue. The red areas show areas in which the FA value (density of white matter) is larger in the experimental group over to control group, and the blue areas show area in which the FA value is larger in the control group. (A larger FA value essentially means that the area in question has more neuronal development.) After we saw that all the data was processed correctly, we started to run the scripts on all the data. We should see those results on Monday.
On Monday, the rest of the summer students arrive. I'm excited to meet all the people I will be working with over the summer. I'm also really happy that I got here early, because I already have a project that I chose, while most of the other summer students are just assigned someone to work with based on a brief summary of their interests. Monday should be a good day!
I also finally downloaded MATLAB today as well as FSL and SPM. Now I can look at brains on my computer! FSL has a very cool way to view the brain: the Orthographic view. When viewing a brain, the orthographic view gives you three windows, each viewing the brain from a different plane (x, y, z. or for anyone who has taken BCS 110, horizontal, sagittal, and coronal.) When you click on any area of the image on any of the planes, the other two images adjust to show that area. It's pretty cool. There is also a 3D view of the brain, but I haven't figured out how to use that yet.
In regards to the DTI research we started running yesterday, I finally got to see some actual data! We finished running all of the scripts we needed to run on the preliminary data, and I got to see the results. (Quick note, the scripts we are running on this data must be run when all the data is present, which it was not yesterday. We did a practice run of the data using only ten of the 27 subjects, so the results we found weren't really results, just pretty cool brain scans.) The experiment is seeing what parts of the brain differ between highly skilled musicians and non-musicians. In the final images, we saw an averaged version of all the data projected onto one "normalized" brain. (I know it sounds weird, but I swear it works) All of the white matter tracks show up as green. Once the experimental conditions are added, some of the green either turns red or blue. The red areas show areas in which the FA value (density of white matter) is larger in the experimental group over to control group, and the blue areas show area in which the FA value is larger in the control group. (A larger FA value essentially means that the area in question has more neuronal development.) After we saw that all the data was processed correctly, we started to run the scripts on all the data. We should see those results on Monday.
On Monday, the rest of the summer students arrive. I'm excited to meet all the people I will be working with over the summer. I'm also really happy that I got here early, because I already have a project that I chose, while most of the other summer students are just assigned someone to work with based on a brief summary of their interests. Monday should be a good day!
I also finally downloaded MATLAB today as well as FSL and SPM. Now I can look at brains on my computer! FSL has a very cool way to view the brain: the Orthographic view. When viewing a brain, the orthographic view gives you three windows, each viewing the brain from a different plane (x, y, z. or for anyone who has taken BCS 110, horizontal, sagittal, and coronal.) When you click on any area of the image on any of the planes, the other two images adjust to show that area. It's pretty cool. There is also a 3D view of the brain, but I haven't figured out how to use that yet.
Thursday, June 3, 2010
Day 3
So I'm just starting this blog on day three of my internship. In the past week, I have possibly learned more than I have in a long time. Last week I was learning UNIX (a computer programming language) and MATLAB (a linear algebra program) so that I could prepare for this internship, and today I was teaching somebody how to use UNIX. Weird.
Today I had my first real fMRI experience! I was a subject in a study about something (no idea what yet). I was in the fMRI machine and the task was an Alcohol Stroop Test. Most people have actually done some form of stroop test without even knowing. Have you ever seen those websites that have names of different colors written out in different colors? (ie green yellow orange purple blue) The task seems pretty simple. In trial 1, you have to read the words, not the colors of the words. However, in the second task, you need to name the colors of the words rather than the words themselves. It's a bit tough. After the Color Stroop test was made, it was expanded into other fields, and this Alcohol Stroop test is a result. In the machine, I was shown series of pictures. In each picture there was an image of either an alcoholic or non-alcoholic beverage, and the name of either an alcoholic or non-alcoholic beverage. In trial one, I had to identify the image as alch or non-alch, while in trial two I had to identify the word. It was a cool study, but wow, that machine is LOUD! The room that the machine is in is freezing cold, so you are in this machine with blankets on top of you, ear plugs in, head taped to a board and secured by cushions. It seems weird that I had to lie still for almost 45 minutes, but honestly, it was nearly impossible to move!
After I got out of the machine, I went back to my desk (yeah, I have a desk... kind of...) and started reading some lectures about how magnetic resonance imaging actually works, specifically fMRI and DTI (Diffusion Tensor Imaging). For the moment, I am (kind of) helping process data in a DTI study involving skilled musicians. DTI tracks the direction in which water molecules diffuse in the brain. These molecules are diffusing along axons (white matter), so by tracking the direction of diffusion, we can see paths of axons in the brain. (DTI is also interesting because it is, I believe, the only type of MRI that tracks white matter. All others track gray matter.) To be totally honest, I am not positive what this study is looking for. I'm going to ask tomorrow. However, what I do know is that the data being used is from 14 highly skilled musicians and 10-16 non-musicians. That's about all I know. What I find interesting, however, is that many of the non-musician's data is coming from older, unrelated experiments. I had no idea that was allowed, but I guess it makes sense. So far we have only moved most of the data into one place (this was when I was teaching somebody UNIX). We should be able to start reading data tomorrow.
Also, I found out today that I get to go into to OR and see two brain surgeries in July! How awesome is that?!
Today I had my first real fMRI experience! I was a subject in a study about something (no idea what yet). I was in the fMRI machine and the task was an Alcohol Stroop Test. Most people have actually done some form of stroop test without even knowing. Have you ever seen those websites that have names of different colors written out in different colors? (ie green yellow orange purple blue) The task seems pretty simple. In trial 1, you have to read the words, not the colors of the words. However, in the second task, you need to name the colors of the words rather than the words themselves. It's a bit tough. After the Color Stroop test was made, it was expanded into other fields, and this Alcohol Stroop test is a result. In the machine, I was shown series of pictures. In each picture there was an image of either an alcoholic or non-alcoholic beverage, and the name of either an alcoholic or non-alcoholic beverage. In trial one, I had to identify the image as alch or non-alch, while in trial two I had to identify the word. It was a cool study, but wow, that machine is LOUD! The room that the machine is in is freezing cold, so you are in this machine with blankets on top of you, ear plugs in, head taped to a board and secured by cushions. It seems weird that I had to lie still for almost 45 minutes, but honestly, it was nearly impossible to move!
After I got out of the machine, I went back to my desk (yeah, I have a desk... kind of...) and started reading some lectures about how magnetic resonance imaging actually works, specifically fMRI and DTI (Diffusion Tensor Imaging). For the moment, I am (kind of) helping process data in a DTI study involving skilled musicians. DTI tracks the direction in which water molecules diffuse in the brain. These molecules are diffusing along axons (white matter), so by tracking the direction of diffusion, we can see paths of axons in the brain. (DTI is also interesting because it is, I believe, the only type of MRI that tracks white matter. All others track gray matter.) To be totally honest, I am not positive what this study is looking for. I'm going to ask tomorrow. However, what I do know is that the data being used is from 14 highly skilled musicians and 10-16 non-musicians. That's about all I know. What I find interesting, however, is that many of the non-musician's data is coming from older, unrelated experiments. I had no idea that was allowed, but I guess it makes sense. So far we have only moved most of the data into one place (this was when I was teaching somebody UNIX). We should be able to start reading data tomorrow.
Also, I found out today that I get to go into to OR and see two brain surgeries in July! How awesome is that?!
Subscribe to:
Comments (Atom)