Readme for Perf_reconstruct_V02.m

******************************************

 This MATLAB function is to reconstruct the raw perfusion images from EPI images by the subtraction between labelled images 
 and control images. Quantified CBF images can also be reconstructed by select the option. It is based on SPM99 and 
 MATLAB(5.3 or above) on Redhat Linux 9 or Windows2000.
    
 All the images are 3D SPM ANALYZE formatted (.img and .hdr). All the results are also saved in SPM ANALYZE format; 
 The labelled and control images should be the data after motion correction.
    
 The method used here are based on the "simple subtraction", "surround subtraction" and "sinc subtraction" approaches described in
 Aguirre GK et al (2002) Experimental design and the relative sensitivity of perfusion and BOLD fMRI, NeuroImage. 15:488-500. 
    
 BOLD data (or whatever the underlying pulse sequence that was used) are generated in addition to the perfusion data
    
 for CASL,
 CBF data are calculated according to the formula from
 Wang J, Alsop DC, et al. (2003) Arterial transit time imaging with flow encoding arterial spin tagging (FEAST).
 Magn Reson Med. 50:599-607. Page600, formula [1]
 CBF_CASL (ml/100g/min) = 60*100*M*¦Ë*R/(2*alp*Mo*(exp(-w*R)-exp(-(t+w)*R))
 where M = raw ASL signal, ¦Ë = blood/tissue water partition coefficient, R =longitudinal relaxation rate of blood,
       alp = tagging efficiency, Mo =  equilibrium magnetization of brain, 
       w = post-labeling delay, t = duration of the labeling pulse,  
 and we use the assumed parameters for calculation as ¦Ë=0.9g/ml, 
 for 3T, alp=0.68, T1b=1490ms, R=1/T1b=0.67sec-1. 
 for 1.5T, alp=0.71, T1b=1200ms, R=1/T1b=0.83sec-1.                                                      

 for PASL,
 CBF data are calculated according to the formula from
 Wang J, Aguirre GK, et al. (2003) Arterial Spin Labeling Perfusion fMRI With Very Low Task Frequency
 Magn Reson Med. 49:796-802. Page798, formula [1]
 CBF_PASL (ml/100g/min) = 60*100*M*¦Ë/(2*alp*Mo*t*exp(-(t+w)*R))
 where M = raw ASL signal, ¦Ë = blood/tissue water partition coefficient, R =longitudinal relaxation rate of blood,
       alp = tagging efficiency, Mo =  equilibrium magnetization of brain, 
       w = post-labeling delay, t = duration of the labeling pulse,  
 and we use the assumed parameters for calculation as ¦Ë=0.9g/ml, 
 for 3T, alp=0.95, T1b=1490ms, R=1/T1b=0.67sec-1. 
 for 1.5T, alp=0.95, T1b=1200ms, R=1/T1b=0.83sec-1.                                                      

******************************************

 function [] = perf_resconstruct(Filename, FieldStrength, ASLType, FirstimageType,...
              SubtractionOrder, SubtractionType, ThreshFlag, threshold, CBFFlag, MeanFlag)

  Inputs:
    Firstimage - integer variable indicating the type of first image 
    - 0:control; 1:labeled 
   Select raw images (*.img, images in a order of control1.img, label1.img, control2.img, label2.img,....;
   or images in a order of label1.img, control1.img, label2.img, control2.img, .... )

    FieldStrength - integer variable indicating the field strenth of the MR scanner
    - 0:1.5T; 1:3T
    
    ASLType - integer variable indicating the type of ASL 
    - 0:Continue ASL; 1:Pulsed ASL
    
    FirstimageType - integer variable indicating whether the first EPI image is labelled 
    - 0:Control (not labelled); 1:Labelled
              
    SubtractionOrder - integer variable indicating the order of subtraction
    - 0:Even-Odd(Img2-Img1); 1:Odd-Even(Img1-Img2)

    SubtractionType - integer variable indicating which subtraction method will be used 
    -0: simple subtraction; 1: surround subtraction;2: sinc subtractioin.
    
    ThreshFlag - integer variable indicating whether Threshold EPI image series 
    - 0:no Threshold; 1:Threshold
    
    threshold - the interger value of threshold the EPI images
    
    CBFFlag - integer variable indicating whether CBF images are produced 
    - 0:no CBF images; 1: produced CBF images
    
    MeanFlag - integer variable indicating whether mean image of all perfusion images are produced 
    - 0:no mean image; 1: produced mean image
    
  Outputs:
    BOld Images: Bold_*.img,Bold_*.hdr;  Mean_Bold.img, Mean_Bold.hdr; 
    Perfusion Images: Perf_*.img, Perf_*.hdr; Mean_Perf.img, Mean_Perf.hdr;
    CBF Images: CBF_*.img, CBF_*.hdr; Mean_CBF.img, Mean_CBF.hdr;
    
    
******************************************
    
  We suggest statistically analyse the CBF data in a way like this (for single subject):   
      
1. Collect the data and tranform all data to SPM ANALYZE format; 

2. Realign all EPI images to do the motion correction; 

3. Set the origin of anatomical T1 image and EPI images;

4. Coregister realigned EPI images to anatomical T1 image;

5. Spatial normalize the anatomical T1 images and write normalized the coregistered EPI images;

6. Spatial smooth the normalized EPI images 

7. run Perf_reconstruct_V02.m, threshold the EPI images and reconstruct Bold, Perf and quantified CBF images and the mean images;

8. specify the design matrix for the analysis of CBF images;
    Attention: Convolve with HRF - "no"

9. estimate the specified fmri model and get the result
    Attention:       High-pass filter? - "none"
                     Low-pass filter? - "none"
    Attention:       one spm file, spm_fmri_spm_ui.m, should be modified 
                         (right-click to download the modified spm_fmri_spm_ui.m file) 
                         Mask Brain? - "no"

******************************************
By Hengyi Rao & Jiongjiong Wang, @CFN, UPenn Med. 07/2004.

if any questions or finding bugs about this program, please send email to hengyi@mail.med.upenn.edu