2010 GUI Interaction E - Inverse Problem Solving
GOAL: Solve inverse problems using Standard Diffusion and Monte Carlo Models.
I. Impact of Optical Properties on Spatially-Resolved Reflectance
- Select the "Inverse Solver Panel".
- For Fwd Model Engine: select "Scaled Monte Carlo - NURBS (g=0.8, n=1.4)", for Inv Model Engine: select "Standard Diffusion (Analytic - Point Source)".
- In Solution Domain: select "Steady State(R(ρ))".
- Set start and stop locations to 0.5 and 9.5 mm, respectively with 10 points (every 1 mm).
- Set Optimization Parameters to: μa and μ's.
- Simulate measured data: set "Forward Simulation Optical Properties:" to: μa = 0.01 mm-1, μ's = 1 mm-1, g = 0.8 and n = 1.4 and 2% noise.
- Confirm the "Hold On" checkbox is checked.
- Click the "Plot Measured Data" button.
- Set "Initial Guess Optical Properties:" to: μa = 0.05 mm-1, μ's = 1.5 mm-1, g = 0.8 and n = 1.4.
- Click the "Plot Initial Guess" button.
- Click the "Run Inverse Solver" button.
- To what optical property values did the inverse solver converge? (Scroll to the bottom of the page to see the output).
- Why are the converged values not exactly the forward simulation optical properties?
- Perform the same analysis with 0% noise added to the simulated measured data. How accurate are the converged properties now?
II. Change Inverse Model Engine
- Perform the same analysis changing the "Inv Model Engine" to Scaled Monte Carlo - NURBS(g=0.8, n=1.4).
- Which Model Engine produced the more accurate converged values?
III. Change number of detectors
- Repeat II using only 2 detectors. Can you strategically place the two detectors to obtain the same accuracy in the converged values as you obtained with 10 detectors?
- Use the analysis plots of ∂R/∂μa and ∂R/∂μ's to help guide their placement.