Kskamp

From REU@MU
Jump to: navigation, search

Comparing Two Models of 99mTc-HMPAO Uptake in the Lungs

Benchmarks

1. Become familiar with MatLab syntax by creating a Computed Tomography (CT) program

2. Create a Graphical User Interface (GUI) to simulate a CT scan

3. Create GUI to simulate forward model of a Single Photon Emission CT (SPECT) scan (with and without noise)

4. Improve SPECT GUI to estimate parameters of the forward model, given the concentration of radioactive material put into the system and output of the noisy lung

5. Determine whether the parameters given by a compartmental model of HMPAO uptake in the lung are reliable

6. Answer the question: Is a Compartmental Model Sufficient to Reliably Estimate Key Physiological Parameters?

Weekly Log

Week 1 (5/31 - 6/3)

- Met with Dr. Clough

   - Discussed scope of project
   - Learned basics of CT, SPECT, and PET scans (projections, Radon transform, Inverse Radon transform, reconstruction)
   - Defined benchmarks for the summer

{ Next meeting Tuesday 6/7 with Dr. Clough and Graduate student(s)}

- Attended Dr. Factor's talk on basics of research

   - Using the tools you have
   - Thoughtful consideration of a problem and solutions
   - Know what you don't know ==> develop a strategy to figure it out
   - Find out what is already known (do you need a new algorithm?)
   - Break the problem down into manageable pieces
   - Know your audience
   

- Conducted background research on CT scans

   - How they have improved (greater efficiency and effectiveness)
   - Confirmed understanding of basics given by Dr. Clough
   

- Wrote preliminary MatLab programs

   - Created image to be scanned
   - Calculated projection image (Radon transform as well as individual projections)
   - Reconstructed image (Inverse radon transform)
   

- Researched basic GUI construction

For next week  : Begin constructing GUI for preliminary CT simulation

Week 2 (6/6 - 6/10)

- Formatted and finished basic CT GUI

   - Fixed image loading issue 
   - Researched solutions for using two axes in one GUI
   - Enhanced GUI (with help from Tony)

- Met with Dr. Clough

   - Introduced to the VA Hospital
   - Toured the lab (CT and SPECT imaging equipment)
   - Introduced to Katherine and Anthony (heretofore referred to as Katie and Tony)
   - Redefined benchmarks for the summer

{ Dr. Clough out of town next Monday-Thursday }

- Watched rat lung experiment measuring mitochondrial productivity before and after the addition of an uncoupler

- Attended Dr. Brylow's talk on basics of research presentation and paper writing

   - What goes into a "good" research talk?
         *Features*
         - Why is your project 'important', 'interesting', and 'hard'?
         - What did YOU do to solve/make progress towards a solution to the problem?
         - Why should they care? (What is the application? ==> sometimes requires creativity)
  
   - What makes a "good" research paper?
         - Answer all the same questions as above
         - Tell the story of the research (X3)
           *tell them what you're going to tell them (Intro)
           *tell them (Content)
           *tell them what you told them (Summary)
         - Tell them why your research is different (Related/Prior work)
   

- Constructed GUI for SPECT forward problem

   - Select data set (in .txt format)
   - Enter values for variables to graph lung curve
   - Save a copy of the GUI interface (including which data file used, all variables and output graph) to pdf

For next week  : Begin work on inverse SPECT GUI

Week 3 (6/13 - 6/17)

- Debugged GUI for SPECT forward problem

   - Emailed Katie code for additional testing

- Formatted and finished SPECT Lung Simulator

   - Fixed lung curve calculation issue 
   - Translated poisson noise code from Ben (C to MatLab)
   - Researched use of Gaussian noise
   - Created additional option of saving simulated lung data points

- Attended Dr. Brylow's presentation on Responsible Conduct of Research

   - OBLIGATIONS
     1) to honor the trust that colleagues place in us
     2) to ourselves and our future careers
     3) to the public
   
   I: ethical treatment of data
      A- how data was obtained
      B- double-blind trials
      C- validated survey
   II: authorship, credit, and plagiarism
      A- fabrication, falsification, and plagiarism
   III: human participants and animal subjects
      A- Common Rule
      B- informed consent
      C- Institutional Research Boards (IRBs)
   IV: intellectual property (who owns your work?)
   V: conflicts of interest and professional standards
      --follow your work to its end

- Conducted background research on HMPAO studies with rats and ALI

- Attended Dr. Corliss' presentation on the basics of LaTeX

   - How to access
   - Basic points of TeX syntax
   - Brief history of TeX and why Knuth created the processor

{ Beginning to work with LaTeX creating research paper }

- Read three papers on background of research on Acute Lung Injury

   - "Differential Lung Uptake of 99mTc-HMPAO and 99mTc-Duramycin in the Chronic Hyperoxia Rat Model"
   - "Role of glutathione in lung retention of 99mTc-HMPAO in two nique rat models of hyperoxic lung injury"
   - "In vivo detection of hyperoxia-induced pulmonary endothelial cell death using 99mTc-Duramycin"

For next week  : Begin work on Inverse Model GUI

Week 4 (6/20 - 6/24)

- Met with Ben

   - Worked out bugs in poisson noise program
   - Fixed error that caused his code not to run on other computers
   - Began reading "Distribution of Capillary Transit Times in Isolated Lungs of Oxygen-Tolerant Rats"
   - Requested a second article from the library

- Met with Dr. Clough

   - Outlined needs for Inverse Problem
      - Non-linear least-squares model regression
      - Input= data (from simulation/measured in lab)
      - Output= values for constants
      - Tests simplified to assume DEM injection (no Glutathione)
   --> Next : use data from PDE model as input
      - Do I get the same constants out?
      - Ratio of calculated constants:actual (known) constants         

- Started solution to inverse problem

   - Read documentation on lsqcurvefit in MatLab
      - Trying to understand how to input vector of parameters into forward model equation
      - Solved the "blackbox" equation problem (variables contained two functions deep in code)
      == added variable to input of forward solver and ODE solver
   - 1st attempts with too many variables (solving for 8 constants)
   - CURRENT WORKING MODEL -
      - Allows computation of up to 3 variables (PS, kOther, kGSHc)
      - Assumes initial conditions of [1, .01, .1], respectively
   - MODIFIED INVERSE MODEL - 
      - Allows computation of up to 2 variables
      - Allows for more direct testing of correlation between two variables
   

- Attended Katie's research talk on the compartmental model

   - Learned a little about modeling, where the model we are working with came from
   - Made notes of things to add to my presentation
   - Listened to questions asked by Dr. Clough and Dr. Audi for more explanation in how her model works
   

- Began writing preliminary Mid-REU presentation

   - Learning to use Beamer on ShareLaTex
   - Working out how much background to give
   - Spoke to Dr. Clough about what to include as part of my talk

- Ran tests with working simulation models

   - Ran forward model to create data for inverse model with known PS, kOther, and kGSHc
   - Tested 16 different trials at 10% error
       - Different trials based on:
            1) changing the initial guess given to the model
            2) shutting off kGSHc pathway (simulating DEM)
            3) shutting off kOther pathway
            4) shutting off PS from the tissue to the blood
   - Retried last 14 trials at 7% error

For Next Week: Finish Presentation, begin practice, evaluate data from Inverse model; read articles from Ben; find an article about sensitivity testing

Week 5 (6/27 - 7/1)

- Met with Dr. Clough

   - Use data from PDE model as input
      - adjusting input program
      - write conversion program for PDE forward data
      - What other modifications need to be made?
   - Differences in single and multiple capillary calculations
   - Gave preliminary talk
      - adjustments based on time constraints
      - reworked description of models
      - corrected visual representation of compartmental model
      - corrected differential equations (and notation)

- Wrote code to extract data from PDE forward output

- Wrote code to format input curve for PDE forward model

- Calculated error based on data from simulation models

   - Calculated each parameter, for both 7% and 10% noise
   - Compiled error calculations into a single sheet to graph        
 

- Finished writing preliminary Mid-REU presentation

   - Made adjustments based on Dr. Clough's recommendations
   - Added graphs of percent error in parameter recovery
   - Wrote out note cards of talking points/explanations
   - Practiced talk

- Gave Mid-REU presentation (and watched others)

For Next Week: Begin comparing PDE data

Week 6 (7/4 - 7/8)

- Rewrote ODE code to calculate mass curves

- Updated Inverse Compartmental model

   - toggles actually have functions
      - mass/concentration curves
      - hold kGSHc/kOther constant (in concentration curve)

- Met with Dr. Clough

   - Recapped talk from Thursday
   - Update of mass curve model
   - New task: calculate equations at steady state
      - PDEs become ODEs wrt Volume (not time)?
      - ODEs become system of algebraic equations
   - Look into Sum of Squared Differences plots

- Dr. Brylow talk: Grad School

   - Do your homework on the schools!
   - TA/RA/Fellowship positions (and differences)
   - Why go to Grad School?
   - Where to look for Grad programs?
   - Who to ask about letters of recommendation?
      - Dr. Krenz's interjection: tell profs in advance if you MIGHT go to them for letters!
   - What does a good essay look like?
   - Cost/benefits of PhD vs Masters--is it worth it for you?
   - Advanced degrees in industry/academia?

- Solved system of ODEs and Algebraic equations

- Wrote code to graph surface of SSD plots

   - trough shaped (PS vs kGSHc)
   - minimum at kGSHc = 0.1 (which was value of given curve)
   - PS didn't have a minimum (can't find it?)

- Met with Dr. Clough

   - Not sure how algebraic system is not dependant on PS?
   - New method for solving ODEs (look for eigenvectors)
   - Compare PDE to ODE data for SSD, PDE to PDE 

- Solved steady state PDEs (ODEs with eigenvalues 0, -(k[12] + k[21]))

For Next Week: Look at PS vs kGSHc for PDE and PDE/ODE

Week 7 (7/11 - 7/15)

- Fixed data extraction code for PDE data

- Wrote Sum of Squared Difference surface function for

   - PDE to ODE (using data extraction code to use PDE data)
   - PDE to PDE (complete function)

- Found and fixed errors in Compartmental model code

- Worked through solution to PDEs and ODEs in steady state with Dr. Clough

   - Why don't they work???

- SSD surfaces == troughs, can we find PS?

   - Got real-world estimated values for PS from Dr. Audi
      - 0<PS<F 
   - New surfaces are triangular troughs
   - Dr. Audi suggested way to define PS
      - Run w/ no uptake in tissue
      - Change flow (and time, and Cin by same relationship)
      - Change in curve=information about PS

- Attended Dr. Factor's talk on posters

   - Less (words) > more (words)
   - Clear, concise, organized
   - Not the WHOLE story--that's why you're there

- Began work on poster

   - Talked to Dr. Clough about "theme"/question
   - Set up organization 
   - Need to save graphs to use
   - FINISH BY MONDAY for review/critique

- Began work on paper

   - Introduction (ALI, HMPAO, hyperoxia rat model)
   - Explanation of models (diagrams and equations)
   - Need to fix constants matrix for compartmental model

For Next Week : Finish poster, run new error calculations for model with PS = .05 (or thereabouts), brainstorm ways to calculate steady state

Week 8 (7/18 - 7/22)

- Calculated error of parameter recovery

   - 7% and 10% error trials
   - Huge error for kGSHc > PS -- WHY?

- Ran tests estimating accuracy of change in parameter

   - How to list the results?
   - How to calculate the results?

- Ran tests of PDE data

   - What causes problems with PDE data?
       - Multiple capillary vs single capillary?
       - Trouble with smooth curve estimations?
   - Very crude parameter estimates

- Proved barrier dependence

   - Read paper as example for how to prove barrier dependence
   - Built program to graph two different flow trials on normalized time

- Submitted final poster

   - Edited poster to include error calculations

- Worked on paper

   - Changed from constants matrix to equation list for compartmental model
   - Wrote preliminary description of procedure followed for research
   - Need to fix equation for compartmental model

For Next Week : Work on paper

Week 9 (7/25 - 7/29)

- Worked on paper

   - Added graphs and captions with descriptions
   - Information of steady state calculations
   - Made changes based on recommendations from Ben and Dr. Clough
   - Fixed constants matrix for compartmental model
   - Wrote conclusion and future works sections
   - Found new sources for paper (rather than quote all from Dr. Clough and Dr. Audi)
   - Changed "recovered" to "estimated" parameters
   - Added indepth description of purpose for SSD graphs

- Wrote abstract for presentation pamphlet

- Presented poster

For Next Week : Work on/finish presentation and paper

Week 10 (8/1 - 8/5)

- Finished presentation

  - Updated slides from mini-presentation
  - Added definitions for PS and kGSHc
  - Added graphs to demonstrate barrier dependence
  - Added information from new error trials

- Edited paper

  - Wrote abstract
  - Made edits on Dr. Clough's recommendation

- Began work on User Guide for MATLAB code and documentation

- Practiced giving presentation

- Gave final presentation

- Completed Post-REU survey for NSF


Personal Information

Kim is studying Mathematics at Lindenwood University in Saint Charles, MO.