README

  ___  ____  __  __
 / __)(  _ \(  \/  )  
 \__ \ )___/ )    (   Statistical Parametric Mapping
 (___/(__)  (_/\/\_)  SPM99 : Updates

________________________________________________________________________
spm99_mexsg64.tar.gz   22/10/2000
------------------------------------------------------------------------

Compilation of mex files for 64 bit SGI machines by Geriant Rees
<geraint@klab.caltech.edu>

________________________________________________________________________
spm_vol_ecat7.m 2.7 00/01/26
------------------------------------------------------------------------
Calibration factor from main hdr is now included in the scalefactor
field of the volume handles. This problem was reported by Steven Grant
<sgrant@nida.nih.gov>.

spm_vol_ecat7.m only used a single scale factor (SCALE_FACTOR) from
the ECAT subheader.  This works fine if and only if the images are
uncalibrated (which they are at the FIL).  If the images are expressed
in specific calibration units (e.g. Bq/cc), then one needs to
incorporate a second factor found in the main header
(mh.ECAT_CALIBRATION_FACTOR).  The product of Scale_Factor and
ECAT_CALIBRATION_FACTOR then gives the proper units.

-JA
________________________________________________________________________
spm_orthviews.m 2.18 00/02/04
------------------------------------------------------------------------
Small bug in display of images.  The first row/column was not being
displayed, although the crosshairs were in the correct positions.
This was noticed by JBP just before the SPM99 release, but there was
no chance to fix it.
-JA

________________________________________________________________________
spm_sn3d.m 2.27 00/02/07
------------------------------------------------------------------------
There is a tiny little bug in batching in spm_sn3d.m that was found by
Matthew Brett.  Line 444... should read:

iterations = spm_input('# Nonlinear Iterations?','+1','m',...
	['1  nonlinear iteration |3  nonlinear iterations|' ...
	 '5  nonlinear iterations|8  nonlinear iterations|' ...
	 '12 nonlinear iterations|16 nonlinear iterations'],...
	 tmp2, tmp,'batch',{},'nonlin_ite_nb');
-JA

________________________________________________________________________
spm_segment.m    2.12 00/02/10
spm_coregister.m 2.10 00/02/10
------------------------------------------------------------------------
Routines modified so that they don't need Graphical displays. This
should allow them to run in batch mode.
-JA

________________________________________________________________________
spm_transverse.m 2.16 00/02/21
spm_mip.m        2.6  00/02/21
spm_mip_ui.m     2.12 00/02/21
------------------------------------------------------------------------
Routines patched so that they work properly for 2D data.  The originals
hadn't been tested.
-JA

________________________________________________________________________
spm_platform.m  2.8 00/02/22
------------------------------------------------------------------------
Matthew Brett's suggestion from 22/02/00:
Alternatively, perhaps it would be better to replace lines 111-126 with:
	twd = tempdir;
-JA

________________________________________________________________________
spm_hwrite.m   2.3 00/02/22
------------------------------------------------------------------------
"spm_hwrite.m" did not like filenames containing spaces, which should
be perfectly commonplace on PCs.  Joseph Dien found this one.
-JA

________________________________________________________________________
spm_coregister.m 2.11 00/02/23
------------------------------------------------------------------------
Numerical instability and underflow problems (which were not causing
warning messages or errors as a matrix pseudo-inverse is used rather
than a proper inverse at line 362 of spm_affsub3.m) solved by modifying
spm_coregister.m at around line 209 so that all the images are scaled
to have roughly the same global intensity.
-JA
________________________________________________________________________
spm_P.m 2.5 00/03/06
------------------------------------------------------------------------
Tom Nichols found an incredibly tiny bug/typo in spm_P.m on line 151:
	c       = (1+t^2/v)^((1-v)/2);
if t is a vector (you pass a vector of statistics to get p-values)
the "t^2" dies with "Matrix must be square".
-JA
________________________________________________________________________
spm_slice_timing.m 2.7 00/03/16
------------------------------------------------------------------------
According to Matthew Brett:
	assuming there is no gap between the end of the
	collection of the last slice, and the beginning
	of the collection of the first slice of the next
	volume, the TA should be specified as
		TA = TR - TR/n

Code modified by Michael Erb:
 line 155: new default value for TA
   old   TA = spm_input('Acquisition Time (TA) {secs}','!+1','e',TR);
   new   TA = spm_input('Acquisition Time (TA) {secs}','!+1','e',TR-TR/nslices);
 line 178: Problem with batch (nargin >= 1)
   old         eval(['P    =    P' num2str(subj) ';']);
   new         if nsubjects > 1, eval(['P    =    P' num2str(subj) ';']); end;
-JA
________________________________________________________________________
render.ZIP
------------------------------------------------------------------------
Sargo Aalto found that the problem often encountered under NT when
loading the render_*.mat files during the Render Display, may have been
due to a problem with Winzip. Matthew Brett suggested that a .zip file
may be useful.  Alternatively, people could dearchive using EasyZip,
which is free from http://www.bamsystems.com/easyzip/ , or pkzip which
is shareware from http://www.pkware.com, or use the GNU tar utility that
is part of the Cygwin project (http://sourceware.cygnus.com/cygwin).
The binary and necessary DLL files are available from Matthew's webspace:
	http://www.mrc-cbu.cam.ac.uk/~matthew.brett/ftp/Win_spm99
Download tar.exe and cygwin1.dll into your path, and, from the Windows DOS
prompt, extract the spm99.tgz file with:
	tar -zxf spm99.tgz
-JA
________________________________________________________________________
spm_sn3d.m 2.28
------------------------------------------------------------------------
Fixed problem reported by Darren Gitelman, in addition to same problem
with bounding box:
| A problem arises in spm_sn3d if one sets the spm global sptl_Vx to 
| empty or 0 OR if one goes into spm_defaults and defines the output 
| voxel sizes to be a runtime option which also sets sptl_Vx to be 0.
| 
| If one then performs a normalization and "Determines Parameters 
| Only," spm_sn3d fails at line 830
| 
| spm_write_sn(P(:,1),matname,bb,Vox,1);
| 
| because Vox is defined to be 0.
| 
| spm_sn3d only asks the user to define the voxel size (as a runtime 
| option) if  either the option write normalized or determine 
| parameters and write normalized were selected. This is because 
| writing out a quick t1 (or object brain) normalization is now a 
| default.
-JA
________________________________________________________________________
spm_read_vols.m 2.5 00/03/21
------------------------------------------------------------------------
Fixed a couple of bugs that Matthew Brett found:
Lines 42, 43 were:
	im = logical(zeros(ni,1));
	for i=1:n, im(i)=~spm_type(Vi(i).dim(4),'NaNrep'); end
but are now:
	im = logical(zeros(n,1));
	for i=1:n, im(i)=~spm_type(V(i).dim(4),'NaNrep'); end
-JA
________________________________________________________________________
spm_help.m 2.43 00/03/23
------------------------------------------------------------------------
Fixed problem pointed out by Jan Kremlacek:
Printing the help to PS, was printed in the BBox bigger than A4 format
even when there is no problem with printing of any results.
-JA
________________________________________________________________________
spm_coreg_ui.m  2.2 00/03/24
------------------------------------------------------------------------
Martin Stumpf noticed a problem with the batch mode version, in that
line 166 has "PFF" for its lvalue rather than "others".
It has now been fixed.
-JA
________________________________________________________________________
spm_get_space.m    2.5  00/03/30
spm_sn3d.m         2.29 00/03/30
spm_bch_bchmat.m   2.7  00/03/30
spm_bch_headers.m  2.6  00/03/30
------------------------------------------------------------------------
A recommendation by Matthew Brett (that he just beat me in finding):

Files modified to save as:
	save('x','y','z');
rather than:
	save x y z
This is because of problems (particularly under Windows) with filenames
and directory paths containing spaces.
-JA
________________________________________________________________________
spm_bch_bchmat.m 2.8 00/05/17
------------------------------------------------------------------------
Bug in version 2.7 found by Petr Janata. The line saying:
	save(spm_str_manip(bch_mfile,'rp'), str)           
was not working since "str" was a string of variable names with
spaces between.
-JA
________________________________________________________________________
spm_list.m       2.31 00/05/25
spm_orthviews.m  2.20 00/05/25
spm_render.m     2.17 00/05/25
spm_transverse.m 2.17 00/05/25
------------------------------------------------------------------------
Problems with negative t values (from conjunction analyses). These were
reported and largely fixed by Darren Gitelman.
-JA
________________________________________________________________________
spm_list_files.c 2.8 00/07/06
------------------------------------------------------------------------
The fix by Volkmar Glauche that resolves the problem with file
permissions under Linux (files get only displayed in spm_get when they
are world-readable). The reason is at line 28 of the original code
(spm_list_files.c)

 	if (uid == -1)

This returns 0 at all times, since under Linux uid is of type unsigned int
and cannot be negative. Replacing this line with the code

 	if (uid == (uid_t)-1)

works with Linux and should not change behaviour on other systems.
-JA
________________________________________________________________________
spm_orthviews.m 2.22 00/07/21
------------------------------------------------------------------------
Included some modifications by Matthew Brett that are to do with
displaying coloured images.  The main reason for inclusion here is
because of the fix for the problem reported by Steve Fromm:

| I don't understand the following behavior:
| 1.  Display the epi.img template, using World Space.  [Say that image
| displayed in Position A]
| 2.  Set yaw = pi/6; image rotates [Position B]
| 3.  Switch to voxel space (leaving yaw set at pi/6); image rotates back to
| the initial position [Position A]
| 4.  Set yaw = 0; image rotates. [Position C; opposite direction, in the
| axial plane]
| 5.  Return to World Space; image rotates back to Position A.
| 6.  Back to Voxel Space; image still in Position A.

What was happening was that any reorienting is undone in the display
when you switch from World->Voxel spaces.
-JA
________________________________________________________________________
spm_sn3d.m 2.30 00/08/07
------------------------------------------------------------------------
Matthew Brett found a truly miniscule bug in spm_sn3d.m.
Line 352 of spm_sn3d.m is missing a quote:
subj(i).objmask = spm_input('batch', {},objmask',i);
should be:
subj(i).objmask = spm_input('batch', {},'objmask',i);
-JA
________________________________________________________________________
spm_graph.m 2.29 00/09/14
------------------------------------------------------------------------
Minor bug in spm_graph.m, affecting plots of event-related fitted
response with PSTH (with no session average). It was found and fixed
by Jorge Armony (ucjtnga@ucl.ac.uk) who replaced line 418
	Y     = Y*ones(length(ss))/length(ss);
with
	Y     = sum(Y,2)/length(ss);
-JA
________________________________________________________________________
spm_transverse.m 2.18 00/10/13
------------------------------------------------------------------------
Minor bug in displaying the titles - discovered by Michele Piana.
The z=?? part for the planes above and below were determined based on
the assumption that the images had been spatially normalised.
-JA
________________________________________________________________________
spm_platform.m  2.10 00/11/08
------------------------------------------------------------------------
Modified (by Volkmar Glauche) to include "GLNX86", which is what Matlab6
reports as the computer platform on Linux boxes
(see spm99_mexglx.tar.gz).
-JA
________________________________________________________________________
spm99_mexglx.tar.gz 00/11/09
------------------------------------------------------------------------
Compilation of mex files for Linux machines for running under Matlab 6.
Files compiled by Volkmar Glauche (see spm_platform.m  2.10).
-JA
________________________________________________________________________
spm_regions.m 2.7	00/11/09
spm_filter.m 2.5 	00/11/09
spm_dcm_display.m 2.1	00/11/09
------------------------------------------------------------------------
spm_regions.m 2.7 allows to extract VOI time-series of adjusted data,
where the VOI can be a sphere, cuboid or a suprathreshold cluster. 
spm_filter 2.5 and spm_dcm_display 2.1 are needed to make spm_regions 2.7
work properly. 
________________________________________________________________________
spm_bch.m        2.8 01/01/29
spm_bch_bchmat.m 2.11 01/01/30
------------------------------------------------------------------------
An assortment of small bug fixes by Darren Gitelman:
* Line 137 of spm_bch.m  should have semicolon at the end of the line.
  index = spm_input('batch',{},'index');

* When one tries to batch defaults_edit the error is in line 165 of 
  spm_bch_bchmat.m
    if exist(bch_names{i})
  exist(bchnames{i}) happens to be true for the bch_name of reset because 
  there is a function reset in toolbox\matlab\graphics\reset.m 

* In spm_bch for the first or second line should include 
  lasterr('');
  this resets the lasterr to null so it doesn't confuse any of the
  try..catch subroutines.
 
* In spm_bch, last line should be clear global BCH. This is so that one 
  could then continue to work in spm normally. If there is a BCH variable 
  then things like dialog boxes don't always show up.
-JA
________________________________________________________________________
spm_mask.m 2.8 01/02/09
------------------------------------------------------------------------
The released version of this function did not work at all (problem
found by Jerry Allison and Emma Burton):
??? Error using ==> run
Undefined function or variable 'p'.

??? Error in ==> /export/home1/seanc/spm99/spm.m
On line 1180  ==> evalin('base',CBs{v-1})

This has been fixed (I think).
-JA
________________________________________________________________________
spm_get.m 2.39 01/02/27
------------------------------------------------------------------------
Included Tom Nichol's fix:
| > I don't know of a good fix... maybe
| > 	
| > 	     'String', cellstr(char(get(findobj(F,'Tag','P'),'UserData')))
| 
| Yup, that fixes it.
________________________________________________________________________
spm_write_plane.m 2.10 00/09/04
------------------------------------------------------------------------
Fixed so that it can be passed data that is not just double precision
floating point.
-JA
________________________________________________________________________
spm_hist2.c 2.3 01/03/09
------------------------------------------------------------------------
Changed it the way the interpolatio is done so that the MI cost
function is less "saw toothed".  Previously, sampling at voxel centres
resulted in higher mutual information than sampling between voxels.

What was happening was that when a point was sampled between the
centres of a group of 8 voxels, a small amount was added to the bins
of the histograms corresponding to the intensity of the 8 neighbours.
The amount added was related to the distance from the point to the
voxel.  When sampling an equal distance from all the points, 1/8 was
added to all the relevent bins.  When sampling exactly at the centre
of a voxel, then 1 was added to the bin for that voxel.  The effect
that this had was that the mutual information cost function was
related to the amount of interpolation done.  When all points were
sampled at the voxel centres, the mutual information was higher
because the histograms were more spikey.  The whole thing resulted in
lots of local minima at positions where there was less interpolation.
It is particularly bad when the images have relatively large voxels.

-JA
________________________________________________________________________
spm_vol_ecat7.m 2.9 01/03/15
spm_vol_minc.m  2.9 01/03/15
------------------------------------------------------------------------
Patched them up for use in Matlab 6. In Matlab5, the following would
work:
	a = struct('b',[]);
	a.b(1)=struct('c','d');
In Matlab 6, the same has to be done something like:
	a = struct('b',[]);
	a.b(1).c = [];

-JA
________________________________________________________________________
spm_grid.m 2.1 01/03/15
------------------------------------------------------------------------
This is a wierd Matlab 6 problem, that appears something to do with
functions that load .mat files, without specifying the names of the
variables to load.
-JA
________________________________________________________________________
------------------------------------------------------------------------