PhantomAnalysisGUI is a tool used for the automated analysis of MRI images of a small structural 
phantom for the assessment of geometric accuracy in preclinical MRI scanners.

The code requires MATLAB version 2015a or later. Unzip all files in a folder and set the MATLAB current
directory to this folder or add it to MATLAB search path.

Instructions for use
--------------------
1. Type 'PhantomAnalysisGUI' in MATLAB's Command Window.
2. Press the 'Load' button on the GUI and select a tif image sequence for analysis (e.g. from the 
   validation dataset).
3. Fill the boxes in the 'Inputs' panel if not filled automatically (for the images in the validation 
   dataset, the voxel size is 0.075 x 0.075 x 1 mm and central slices 7 to 11 are required for analysis).
4. Press the 'Measure dimensions' or the 'Measure volume' button in the 'Analysis' panel to measure the 
   phantom's dimensions or the volume of its central cylindrical compartment respectively. 

Notes:
- Canny edge detection was found to be the most accurate method for these analyses and is set as the 
  default, but k-means can be used instead. Choose a method in the 'Analysis' panel by pressing the
  appropriate buttons and change automatically estimated Canny parameters or choose 2 or 3 clusters for 
  k-means in the 'Inputs' panel.
- Regarding the measurement of dimensions, increasing the scale factor for sub-pixel analysis will 
  compromise the robustness and speed of analysis. A factor of 5 is the default, meaning that the 
  precision of measurements in any direction is 5 times higher than the voxel size in that direction.
- The GUI's log window lists all the measurements performed thus far. Dimensions and volume estimates 
  per slice are shown in MATLAB's Command Window.
- The analysis is visualised in the 'Output Data' figure. Select 'Show: Overlaid' to show the phantom 
  boundary over the corresponding slices, or 'Show: Masks only' to show the phantom masks. The red
  lines represent the detected edges and the small green cross marks the positions where dimensional 
  measurements are performed; the distance between opposing cross marks is the measured dimension. 
- Check 'Save images?' and press 'Measure dimensions' or 'Measure volume' again in the 'Analysis' panel 
  to save images showing the analysis process and final phantom masks in a new folder in the directory 
  where the input image sequence is.

Validation dataset
------------------
The validation dataset was created by randomly scaling and translating actual phantom scans and then 
randomly adding (Rician distribution) or removing noise (median filters). The dataset includes:
- scans acquired using a surface (01-30) and a volume coil (31-60)
- scans in the axial (01-05, 31-35), axial 90 (06-10, 36-40), coronal (11-15, 41-45), 
  coronal 90 (16-20, 46-50), sagittal (21-25, 51-55) and sagittal 90 plane (26-30, 56-60)

All scans were used for validating the tool's accuracy of dimension measurement, and axial scans alone
for validating the tool's accuracy of volume measurement. Although the tool is designed to be able to
measure the volume in scans in the other planes as well, it might have a higher chance of error. For 
both analyses, axial 90 and sagittal scans are rotated 90 degrees clockwise.


by Xenios Milidonis, last modified 11 August 2016
email: x.milidonis@sms.ed.ac.uk