Пример #1
0
int main( int argc , char * argv[] )
{
   int do_norm=0 , qdet=2 , have_freq=0 , do_automask=0 ;
   float dt=0.0f , fbot=0.0f,ftop=999999.9f , blur=0.0f ;
   MRI_IMARR *ortar=NULL ; MRI_IMAGE *ortim=NULL ;
   THD_3dim_dataset **ortset=NULL ; int nortset=0 ;
   THD_3dim_dataset *inset=NULL , *outset=NULL;
   char *prefix="RSFC" ;
   byte *mask=NULL ;
   int mask_nx=0,mask_ny=0,mask_nz=0,nmask , verb=1 , 
		nx,ny,nz,nvox , nfft=0 , kk ;
   float **vec , **ort=NULL ; int nort=0 , vv , nopt , ntime  ;
   MRI_vectim *mrv ;
   float pvrad=0.0f ; int nosat=0 ;
   int do_despike=0 ;

	// @@ non-BP variables
	float fbotALL=0.0f, ftopALL=999999.9f; // do full range version
	int NumDen = 0; // switch for doing numerator or denom
	THD_3dim_dataset *outsetALL=NULL ; 	
	int m, mm;
	float delf; // harmonics
	int ind_low,ind_high,N_ny, ctr;
	float sqnt,nt_fac;
	gsl_fft_real_wavetable *real1, *real2; // GSL stuff
	gsl_fft_real_workspace *work;
	double *series1, *series2;	
	double *xx1,*xx2;
	float numer,denom,val;
	float *alff=NULL,*malff=NULL,*falff=NULL,
         *rsfa=NULL,*mrsfa=NULL,*frsfa=NULL; // values
	float meanALFF=0.0f,meanRSFA=0.0f; // will be for mean in brain region
	THD_3dim_dataset *outsetALFF=NULL;
	THD_3dim_dataset *outsetmALFF=NULL;
	THD_3dim_dataset *outsetfALFF=NULL;
	THD_3dim_dataset *outsetRSFA=NULL;
	THD_3dim_dataset *outsetmRSFA=NULL;
	THD_3dim_dataset *outsetfRSFA=NULL;
	char out_lff[300];
	char out_alff[300];
	char out_malff[300];
	char out_falff[300];
	char out_rsfa[300];
	char out_mrsfa[300];
	char out_frsfa[300];
	char out_unBP[300];
	int SERIES_OUT = 1;
	int UNBP_OUT = 0; 
	int DO_RSFA = 1;
	int BP_LAST = 0; // option for only doing filter to LFFs at very end of proc
	float de_rsfa=0.0f,nu_rsfa=0.0f;
	double pow1=0.0,pow2=0.0;

   /*-- help? --*/

   if( argc < 2 || strcmp(argv[1],"-help") == 0 ){
		printf(
"\n  Program to calculate common resting state functional connectivity (RSFC)\n"
"  parameters (ALFF, mALFF, fALFF, RSFA, etc.) for resting state time\n"
"  series.  This program is **heavily** based on the existing\n"
"  3dBandPass by RW Cox, with the amendments to calculate RSFC\n"
"  parameters written by PA Taylor (July, 2012).\n"
"  This program is part of FATCAT (Taylor & Saad, 2013) in AFNI. Importantly,\n"
"  its functionality can be included in the `afni_proc.py' processing-script \n"
"  generator; see that program's help file for an example including RSFC\n"
"  and spectral parameter calculation via the `-regress_RSFC' option.\n"
"\n"
"* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\n"
"\n"
"  All options of 3dBandPass may be used here (with a couple other\n"
"  parameter options, as well): essentially, the motivation of this\n"
"  program is to produce ALFF, etc. values of the actual RSFC time\n"
"  series that you calculate.  Therefore, all the 3dBandPass processing\n"
"  you normally do en route to making your final `resting state time\n"
"  series' is done here to generate your LFFs, from which the\n"
"  amplitudes in the LFF band are calculated at the end.  In order to\n"
"  calculate fALFF, the same initial time series are put through the\n"
"  same processing steps which you have chosen but *without* the\n"
"  bandpass part; the spectrum of this second time series is used to\n"
"  calculate the fALFF denominator.\n"
" \n"
"  For more information about each RSFC parameter, see, e.g.:   \n"
"  ALFF/mALFF -- Zang et al. (2007),\n"
"  fALFF --      Zou et al. (2008),\n"
"  RSFA --       Kannurpatti & Biswal (2008).\n"
"\n"
"* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\n"
"\n"
" + USAGE: 3dRSFC [options] fbot ftop dataset\n"
"\n"
"* One function of this program is to prepare datasets for input\n"
"   to 3dSetupGroupInCorr.  Other uses are left to your imagination.\n"
"\n"
"* 'dataset' is a 3D+time sequence of volumes\n"
"   ++ This must be a single imaging run -- that is, no discontinuities\n"
"       in time from 3dTcat-ing multiple datasets together.\n"
"\n"
"* fbot = lowest frequency in the passband, in Hz\n"
"   ++ fbot can be 0 if you want to do a lowpass filter only;\n"
"       HOWEVER, the mean and Nyquist freq are always removed.\n"
"\n"
"* ftop = highest frequency in the passband (must be > fbot)\n"
"   ++ if ftop > Nyquist freq, then it's a highpass filter only.\n"
"\n"
"* Set fbot=0 and ftop=99999 to do an 'allpass' filter.\n"
"  ++ Except for removal of the 0 and Nyquist frequencies, that is.\n"
"\n"
"* You cannot construct a 'notch' filter with this program!\n"
"  ++ You could use 3dRSFC followed by 3dcalc to get the same effect.\n"
"  ++ If you are understand what you are doing, that is.\n"
"  ++ Of course, that is the AFNI way -- if you don't want to\n"
"     understand what you are doing, use Some other PrograM, and\n"
"     you can still get Fine StatisticaL maps.\n"
"\n"
"* 3dRSFC will fail if fbot and ftop are too close for comfort.\n"
"  ++ Which means closer than one frequency grid step df,\n"
"     where df = 1 / (nfft * dt) [of course]\n"
"\n"
"* The actual FFT length used will be printed, and may be larger\n"
"   than the input time series length for the sake of efficiency.\n"
"  ++ The program will use a power-of-2, possibly multiplied by\n"
"     a power of 3 and/or 5 (up to and including the 3rd power of\n"
"     each of these: 3, 9, 27, and 5, 25, 125).\n"
"\n"
"* Note that the results of combining 3dDetrend and 3dRSFC will\n"
"   depend on the order in which you run these programs.  That's why\n"
"   3dRSFC has the '-ort' and '-dsort' options, so that the\n"
"   time series filtering can be done properly, in one place.\n"
"\n"
"* The output dataset is stored in float format.\n"
"\n"
"* The order of processing steps is the following (most are optional), and\n"
"  for the LFFs, the bandpass is done between the specified fbot and ftop,\n"
"  while for the `whole spectrum' (i.e., fALFF denominator) the bandpass is:\n"
"  done only to exclude the time series mean and the Nyquist frequency:\n"
" (0) Check time series for initial transients [does not alter data]\n"
" (1) Despiking of each time series\n"
" (2) Removal of a constant+linear+quadratic trend in each time series\n"
" (3) Bandpass of data time series\n"
" (4) Bandpass of -ort time series, then detrending of data\n"
"      with respect to the -ort time series\n"
" (5) Bandpass and de-orting of the -dsort dataset,\n"
"      then detrending of the data with respect to -dsort\n"
" (6) Blurring inside the mask [might be slow]\n"
" (7) Local PV calculation     [WILL be slow!]\n"
" (8) L2 normalization         [will be fast.]\n"
" (9) Calculate spectrum and amplitudes, for RSFC parameters.\n"
"\n"
"* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\n"
"--------\n"
"OPTIONS:\n"
"--------\n"
" -despike        = Despike each time series before other processing.\n"
"                   ++ Hopefully, you don't actually need to do this,\n"
"                      which is why it is optional.\n"
" -ort f.1D       = Also orthogonalize input to columns in f.1D\n"
"                   ++ Multiple '-ort' options are allowed.\n"
" -dsort fset     = Orthogonalize each voxel to the corresponding\n"
"                    voxel time series in dataset 'fset', which must\n"
"                    have the same spatial and temporal grid structure\n"
"                    as the main input dataset.\n"
"                   ++ At present, only one '-dsort' option is allowed.\n"
" -nodetrend      = Skip the quadratic detrending of the input that\n"
"                    occurs before the FFT-based bandpassing.\n"
"                   ++ You would only want to do this if the dataset\n"
"                      had been detrended already in some other program.\n"
" -dt dd          = set time step to 'dd' sec [default=from dataset header]\n"
" -nfft N         = set the FFT length to 'N' [must be a legal value]\n"
" -norm           = Make all output time series have L2 norm = 1\n"
"                   ++ i.e., sum of squares = 1\n"
" -mask mset      = Mask dataset\n"
" -automask       = Create a mask from the input dataset\n"
" -blur fff       = Blur (inside the mask only) with a filter\n"
"                    width (FWHM) of 'fff' millimeters.\n"
" -localPV rrr    = Replace each vector by the local Principal Vector\n"
"                    (AKA first singular vector) from a neighborhood\n"
"                    of radius 'rrr' millimiters.\n"
"                   ++ Note that the PV time series is L2 normalized.\n"
"                   ++ This option is mostly for Bob Cox to have fun with.\n"
"\n"
" -input dataset  = Alternative way to specify input dataset.\n"
" -band fbot ftop = Alternative way to specify passband frequencies.\n"
"\n"
" -prefix ppp     = Set prefix name of output dataset. Name of filtered time\n"
"                   series would be, e.g., ppp_LFF+orig.*, and the parameter\n"
"                   outputs are named with obvious suffices.\n"
" -quiet          = Turn off the fun and informative messages. (Why?)\n"
" -no_rs_out      = Don't output processed time series-- just output\n"
"                   parameters (not recommended, since the point of\n"
"                   calculating RSFC params here is to have them be quite\n"
"                   related to the time series themselves which are used for\n"
"                   further analysis)."
" -un_bp_out      = Output the un-bandpassed series as well (default is not \n"
"                   to).  Name would be, e.g., ppp_unBP+orig.* .\n"
"                   with suffix `_unBP'.\n"
" -no_rsfa        = If you don't want RSFA output (default is to do so).\n"
" -bp_at_end      = A (probably unnecessary) switch to have bandpassing be \n"
"                   the very last processing step that is done in the\n"
"                   sequence of steps listed above; at Step 3 above, only \n"
"                   the time series mean and nyquist are BP'ed out, and then\n"
"                   the LFF series is created only after Step 9.  NB: this \n"
"                   probably makes only very small changes for most\n"
"                   processing sequences (but maybe not, depending usage).\n"
"\n"
" -notrans        = Don't check for initial positive transients in the data:\n"
"  *OR*             ++ The test is a little slow, so skipping it is OK,\n"
" -nosat               if you KNOW the data time series are transient-free.\n"
"                   ++ Or set AFNI_SKIP_SATCHECK to YES.\n"
"                   ++ Initial transients won't be handled well by the\n"
"                      bandpassing algorithm, and in addition may seriously\n"
"                      contaminate any further processing, such as inter-\n"
"                      voxel correlations via InstaCorr.\n"
"                   ++ No other tests are made [yet] for non-stationary \n"
"                      behavior in the time series data.\n"
"\n"
"* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\n"
"\n"
"  If you use this program, please reference the introductory/description\n"
"  paper for the FATCAT toolbox:\n"
"        Taylor PA, Saad ZS (2013).  FATCAT: (An Efficient) Functional\n"
"        And Tractographic Connectivity Analysis Toolbox. Brain \n"
"        Connectivity 3(5):523-535.\n"
"____________________________________________________________________________\n"
);
		PRINT_AFNI_OMP_USAGE(
" 3dRSFC" ,
" * At present, the only part of 3dRSFC that is parallelized is the\n"
"   '-blur' option, which processes each sub-brick independently.\n"
									) ;
		PRINT_COMPILE_DATE ; exit(0) ;
   }
	
   /*-- startup --*/
	
   mainENTRY("3dRSFC"); machdep();
   AFNI_logger("3dRSFC",argc,argv);
   PRINT_VERSION("3dRSFC (from 3dBandpass by RW Cox): version THETA"); 
	AUTHOR("PA Taylor");
	
   nosat =  AFNI_yesenv("AFNI_SKIP_SATCHECK") ;
	
   nopt = 1 ;
   while( nopt < argc && argv[nopt][0] == '-' ){

		if( strcmp(argv[nopt],"-despike") == 0 ){  /* 08 Oct 2010 */
			do_despike++ ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-nfft") == 0 ){
			int nnup ;
			if( ++nopt >= argc ) ERROR_exit("need an argument after -nfft!") ;
			nfft = (int)strtod(argv[nopt],NULL) ;
			nnup = csfft_nextup_even(nfft) ;
			if( nfft < 16 || nfft != nnup )
				ERROR_exit("value %d after -nfft is illegal! Next legal value = %d",nfft,nnup) ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-blur") == 0 ){
			if( ++nopt >= argc ) ERROR_exit("need an argument after -blur!") ;
			blur = strtod(argv[nopt],NULL) ;
			if( blur <= 0.0f ) WARNING_message("non-positive blur?!") ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-localPV") == 0 ){
			if( ++nopt >= argc ) ERROR_exit("need an argument after -localpv!") ;
			pvrad = strtod(argv[nopt],NULL) ;
			if( pvrad <= 0.0f ) WARNING_message("non-positive -localpv?!") ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-prefix") == 0 ){
			if( ++nopt >= argc ) ERROR_exit("need an argument after -prefix!") ;
			prefix = strdup(argv[nopt]) ;
			if( !THD_filename_ok(prefix) ) ERROR_exit("bad -prefix option!") ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-automask") == 0 ){
			if( mask != NULL ) ERROR_exit("Can't use -mask AND -automask!") ;
			do_automask = 1 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-mask") == 0 ){
			THD_3dim_dataset *mset ;
			if( ++nopt >= argc ) ERROR_exit("Need argument after '-mask'") ;
			if( mask != NULL || do_automask ) ERROR_exit("Can't have two mask inputs") ;
			mset = THD_open_dataset( argv[nopt] ) ;
			CHECK_OPEN_ERROR(mset,argv[nopt]) ;
			DSET_load(mset) ; CHECK_LOAD_ERROR(mset) ;
			mask_nx = DSET_NX(mset); mask_ny = DSET_NY(mset); mask_nz = DSET_NZ(mset);
			mask = THD_makemask( mset , 0 , 0.5f, 0.0f ) ; DSET_delete(mset) ;
			if( mask == NULL ) ERROR_exit("Can't make mask from dataset '%s'",argv[nopt]) ;
			nmask = THD_countmask( mask_nx*mask_ny*mask_nz , mask ) ;
			if( verb ) INFO_message("Number of voxels in mask = %d",nmask) ;
			if( nmask < 1 ) ERROR_exit("Mask is too small to process") ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-norm") == 0 ){
			do_norm = 1 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-quiet") == 0 ){
			verb = 0 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-no_rs_out") == 0 ){ // @@
			SERIES_OUT = 0 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-un_bp_out") == 0 ){ // @@
			UNBP_OUT = 1 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-no_rsfa") == 0 ){ // @@
			DO_RSFA = 0 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-bp_at_end") == 0 ){ // @@
			BP_LAST = 1 ; nopt++ ; continue ;
		}




		if( strcmp(argv[nopt],"-notrans") == 0 || strcmp(argv[nopt],"-nosat") == 0 ){
			nosat = 1 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-ort") == 0 ){
			if( ++nopt >= argc ) ERROR_exit("need an argument after -ort!") ;
			if( ortar == NULL ) INIT_IMARR(ortar) ;
			ortim = mri_read_1D( argv[nopt] ) ;
			if( ortim == NULL ) ERROR_exit("can't read from -ort '%s'",argv[nopt]) ;
			mri_add_name(argv[nopt],ortim) ;
			ADDTO_IMARR(ortar,ortim) ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-dsort") == 0 ){
			THD_3dim_dataset *qset ;
			if( ++nopt >= argc ) ERROR_exit("need an argument after -dsort!") ;
			if( nortset > 0 ) ERROR_exit("only 1 -dsort option is allowed!") ;
			qset = THD_open_dataset(argv[nopt]) ;
			CHECK_OPEN_ERROR(qset,argv[nopt]) ;
			ortset = (THD_3dim_dataset **)realloc(ortset,
															  sizeof(THD_3dim_dataset *)*(nortset+1)) ;
			ortset[nortset++] = qset ;
			nopt++ ; continue ;
		}

		if( strncmp(argv[nopt],"-nodetrend",6) == 0 ){
			qdet = 0 ; nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-dt") == 0 ){
			if( ++nopt >= argc ) ERROR_exit("need an argument after -dt!") ;
			dt = (float)strtod(argv[nopt],NULL) ;
			if( dt <= 0.0f ) WARNING_message("value after -dt illegal!") ;
			nopt++ ; continue ;
		}

		if( strcmp(argv[nopt],"-input") == 0 ){
			if( inset != NULL ) ERROR_exit("Can't have 2 -input options!") ;
			if( ++nopt >= argc ) ERROR_exit("need an argument after -input!") ;
			inset = THD_open_dataset(argv[nopt]) ;
			CHECK_OPEN_ERROR(inset,argv[nopt]) ; 

			nopt++ ; continue ;
		}

		if( strncmp(argv[nopt],"-band",5) == 0 ){
			if( ++nopt >= argc-1 ) ERROR_exit("need 2 arguments after -band!") ;
			if( have_freq ) WARNING_message("second -band option replaces first one!") ;
			fbot = strtod(argv[nopt++],NULL) ;
			ftop = strtod(argv[nopt++],NULL) ;
			have_freq = 1 ; continue ;
		}

		ERROR_exit("Unknown option: '%s'",argv[nopt]) ;
   }

   /** check inputs for reasonablositiness **/

   if( !have_freq ){
		if( nopt+1 >= argc )
			ERROR_exit("Need frequencies on command line after options!") ;
		fbot = (float)strtod(argv[nopt++],NULL) ;
		ftop = (float)strtod(argv[nopt++],NULL) ;
   }

   if( inset == NULL ){
		if( nopt >= argc )
			ERROR_exit("Need input dataset name on command line after options!") ;
		inset = THD_open_dataset(argv[nopt]) ;
		CHECK_OPEN_ERROR(inset,argv[nopt]) ;	 

		nopt++ ;
   }
   DSET_UNMSEC(inset) ;

   if( fbot < 0.0f  ) ERROR_exit("fbot value can't be negative!") ;
   if( ftop <= fbot ) ERROR_exit("ftop value %g must be greater than fbot value %g!",ftop,fbot) ;

   ntime = DSET_NVALS(inset) ;
   if( ntime < 9 ) ERROR_exit("Input dataset is too short!") ;

   if( nfft <= 0 ){
		nfft = csfft_nextup_even(ntime) ;
		if( verb ) INFO_message("Data length = %d  FFT length = %d",ntime,nfft) ;
		(void)THD_bandpass_set_nfft(nfft) ;
   } else if( nfft < ntime ){
		ERROR_exit("-nfft %d is less than data length = %d",nfft,ntime) ;
   } else {
		kk = THD_bandpass_set_nfft(nfft) ;
		if( kk != nfft && verb )
			INFO_message("Data length = %d  FFT length = %d",ntime,kk) ;
   }

   if( dt <= 0.0f ){
		dt = DSET_TR(inset) ;
		if( dt <= 0.0f ){
			WARNING_message("Setting dt=1.0 since input dataset lacks a time axis!") ;
			dt = 1.0f ;
		}
   }
   ftopALL = 1./dt ;// Aug,2016: should solve problem of a too-large
                    // value for THD_bandpass_vectors(), while still
                    // being >f_{Nyquist}

   if( !THD_bandpass_OK(ntime,dt,fbot,ftop,1) ) ERROR_exit("Can't continue!") ;

   nx = DSET_NX(inset); ny = DSET_NY(inset); nz = DSET_NZ(inset); nvox = nx*ny*nz;

   /* check mask, or create it */

   if( verb ) INFO_message("Loading input dataset time series" ) ;
   DSET_load(inset) ;

   if( mask != NULL ){
		if( mask_nx != nx || mask_ny != ny || mask_nz != nz )
			ERROR_exit("-mask dataset grid doesn't match input dataset") ;

   } else if( do_automask ){
		mask = THD_automask( inset ) ;
		if( mask == NULL )
			ERROR_message("Can't create -automask from input dataset?") ;
		nmask = THD_countmask( DSET_NVOX(inset) , mask ) ;
		if( verb ) INFO_message("Number of voxels in automask = %d",nmask);
		if( nmask < 1 ) ERROR_exit("Automask is too small to process") ;

   } else {
		mask = (byte *)malloc(sizeof(byte)*nvox) ; nmask = nvox ;
		memset(mask,1,sizeof(byte)*nvox) ;
		// if( verb ) // @@ alert if aaaalllllll vox are going to be analyzed!
		INFO_message("No mask ==> processing all %d voxels",nvox);
   }

   /* A simple check of dataset quality [08 Feb 2010] */

   if( !nosat ){
		float val ;
		INFO_message(
						 "Checking dataset for initial transients [use '-notrans' to skip this test]") ;
		val = THD_saturation_check(inset,mask,0,0) ; kk = (int)(val+0.54321f) ;
		if( kk > 0 )
			ININFO_message(
								"Looks like there %s %d non-steady-state initial time point%s :-(" ,
								((kk==1) ? "is" : "are") , kk , ((kk==1) ? " " : "s") ) ;
		else if( val > 0.3210f )  /* don't ask where this threshold comes from! */
			ININFO_message(
								"MAYBE there's an initial positive transient of 1 point, but it's hard to tell\n") ;
		else
			ININFO_message("No widespread initial positive transient detected :-)") ;
   }

   /* check -dsort inputs for match to inset */

   for( kk=0 ; kk < nortset ; kk++ ){
		if( DSET_NX(ortset[kk])    != nx ||
			 DSET_NY(ortset[kk])    != ny ||
			 DSET_NZ(ortset[kk])    != nz ||
			 DSET_NVALS(ortset[kk]) != ntime )
			ERROR_exit("-dsort %s doesn't match input dataset grid" ,
						  DSET_BRIKNAME(ortset[kk]) ) ;
   }

   /* convert input dataset to a vectim, which is more fun */

	// @@ convert BP'ing ftop/bot into indices for the DFT (below)
	delf = 1.0/(ntime*dt); 
	ind_low = (int) rint(fbot/delf);
	ind_high = (int) rint(ftop/delf);
	if( ntime % 2 ) // nyquist number
		N_ny = (ntime-1)/2;
	else
		N_ny = ntime/2;
	sqnt = sqrt(ntime);
	nt_fac = sqrt(ntime*(ntime-1));

	// @@ if BP_LAST==0:
	// now we go through twice, doing LFF bandpass for NumDen==0 and
	// `full spectrum' processing for NumDen==1.
	// if BP_LAST==1:
	// now we go through once, doing only `full spectrum' processing
	for( NumDen=0 ; NumDen<2 ; NumDen++) {
		//if( NumDen==1 ){ // full spectrum
		//	fbot = fbotALL;
		//	ftop = ftopALL;
		//}
		
		// essentially, just doesn't BP here, and the perfect filtering at end
		// is used for both still; this makes the final output spectrum
		// contain only frequencies in range of 0.01-0.08
		if( BP_LAST==1 )
			INFO_message("Only doing filtering to LFFs at end!");
		
		
		mrv = THD_dset_to_vectim( inset , mask , 0 ) ;
		if( mrv == NULL ) ERROR_exit("Can't load time series data!?") ;
		if( NumDen==1 )
			DSET_unload(inset) ; // @@ only unload on 2nd pass

		/* similarly for the ort vectors */

		if( ortar != NULL ){
			for( kk=0 ; kk < IMARR_COUNT(ortar) ; kk++ ){
				ortim = IMARR_SUBIM(ortar,kk) ;
				if( ortim->nx < ntime )
					ERROR_exit("-ort file %s is shorter than input dataset time series",
								  ortim->name ) ;
				ort  = (float **)realloc( ort , sizeof(float *)*(nort+ortim->ny) ) ;
				for( vv=0 ; vv < ortim->ny ; vv++ )
					ort[nort++] = MRI_FLOAT_PTR(ortim) + ortim->nx * vv ;
			}
		}

		/* all the real work now */

		if( do_despike ){
			int_pair nsp ;
			if( verb ) INFO_message("Testing data time series for spikes") ;
			nsp = THD_vectim_despike9( mrv ) ;
			if( verb ) ININFO_message(" -- Squashed %d spikes from %d voxels",nsp.j,nsp.i) ;
		}

		if( verb ) INFO_message("Bandpassing data time series") ;

		if( (BP_LAST==0) && (NumDen==0) )
			(void)THD_bandpass_vectim( mrv , dt,fbot,ftop , qdet , nort,ort ) ;
		else
			(void)THD_bandpass_vectim( mrv , dt,fbotALL,ftopALL, qdet,nort,ort ) ;

		/* OK, maybe a little more work */

		if( nortset == 1 ){
			MRI_vectim *orv ;
			orv = THD_dset_to_vectim( ortset[0] , mask , 0 ) ;
			if( orv == NULL ){
				ERROR_message("Can't load -dsort %s",DSET_BRIKNAME(ortset[0])) ;
			} else {
				float *dp , *mvv , *ovv , ff ;
				if( verb ) INFO_message("Orthogonalizing to bandpassed -dsort") ;
				//(void)THD_bandpass_vectim( orv , dt,fbot,ftop , qdet , nort,ort ) ; //@@
				if( (BP_LAST==0) && (NumDen==0) )
					(void)THD_bandpass_vectim(orv,dt,fbot,ftop,qdet,nort,ort);
				else
					(void)THD_bandpass_vectim(orv,dt,fbotALL,ftopALL,qdet,nort,ort);

				THD_vectim_normalize( orv ) ;
				dp = malloc(sizeof(float)*mrv->nvec) ;
				THD_vectim_vectim_dot( mrv , orv , dp ) ;
				for( vv=0 ; vv < mrv->nvec ; vv++ ){
					ff = dp[vv] ;
					if( ff != 0.0f ){
						mvv = VECTIM_PTR(mrv,vv) ; ovv = VECTIM_PTR(orv,vv) ;
						for( kk=0 ; kk < ntime ; kk++ ) mvv[kk] -= ff*ovv[kk] ;
					}
				}
				VECTIM_destroy(orv) ; free(dp) ;
			}
		}

		if( blur > 0.0f ){
			if( verb )
				INFO_message("Blurring time series data spatially; FWHM=%.2f",blur) ;
			mri_blur3D_vectim( mrv , blur ) ;
		}
		if( pvrad > 0.0f ){
			if( verb )
				INFO_message("Local PV-ing time series data spatially; radius=%.2f",pvrad) ;
			THD_vectim_normalize( mrv ) ;
			THD_vectim_localpv( mrv , pvrad ) ;
		}
		if( do_norm && pvrad <= 0.0f ){
			if( verb ) INFO_message("L2 normalizing time series data") ;
			THD_vectim_normalize( mrv ) ;
		}

		/* create output dataset, populate it, write it, then quit */
		if( (NumDen==0) ) { // @@ BP'ed version;  will do filt if BP_LAST

			if(BP_LAST) // do bandpass here for BP_LAST
				(void)THD_bandpass_vectim(mrv,dt,fbot,ftop,qdet,0,NULL);

			if( verb ) INFO_message("Creating output dataset in memory, then writing it") ;
			outset = EDIT_empty_copy(inset) ;
			if(SERIES_OUT){
				sprintf(out_lff,"%s_LFF",prefix); 
				EDIT_dset_items( outset , ADN_prefix,out_lff , ADN_none ) ;
				tross_Copy_History( inset , outset ) ;
				tross_Make_History( "3dBandpass" , argc,argv , outset ) ;
			}
			for( vv=0 ; vv < ntime ; vv++ )
				EDIT_substitute_brick( outset , vv , MRI_float , NULL ) ;
		
#if 1
			THD_vectim_to_dset( mrv , outset ) ;
#else
			AFNI_OMP_START ;
#pragma omp parallel
			{ float *far , *var ; int *ivec=mrv->ivec ; int vv,kk ;
#pragma omp for
				for( vv=0 ; vv < ntime ; vv++ ){
					far = DSET_BRICK_ARRAY(outset,vv) ; var = mrv->fvec + vv ;
					for( kk=0 ; kk < nmask ; kk++ ) far[ivec[kk]] = var[kk*ntime] ;
				}
			}
			AFNI_OMP_END ;
#endif
			VECTIM_destroy(mrv) ;
			if(SERIES_OUT){ // @@
				DSET_write(outset) ; if( verb ) WROTE_DSET(outset) ;
			}
		}
		else{ // @@ non-BP'ed version
			if( verb ) INFO_message("Creating output dataset 2 in memory") ;

			// do this here because LFF version was also BP'ed at end.
			if(BP_LAST) // do bandpass here for BP_LAST
				(void)THD_bandpass_vectim(mrv,dt,fbotALL,ftopALL,qdet,0,NULL);

			outsetALL = EDIT_empty_copy(inset) ;
			if(UNBP_OUT){ 
				sprintf(out_unBP,"%s_unBP",prefix); 
				EDIT_dset_items( outsetALL, ADN_prefix, out_unBP, ADN_none );
				tross_Copy_History( inset , outsetALL ) ;
				tross_Make_History( "3dRSFC" , argc,argv , outsetALL ) ;
			}
			for( vv=0 ; vv < ntime ; vv++ )
				EDIT_substitute_brick( outsetALL , vv , MRI_float , NULL ) ;
		
#if 1
			THD_vectim_to_dset( mrv , outsetALL ) ;
#else
			AFNI_OMP_START ;
#pragma omp parallel
			{ float *far , *var ; int *ivec=mrv->ivec ; int vv,kk ;
#pragma omp for
				for( vv=0 ; vv < ntime ; vv++ ){
					far = DSET_BRICK_ARRAY(outsetALL,vv) ; var = mrv->fvec + vv ;
					for( kk=0 ; kk < nmask ; kk++ ) far[ivec[kk]] = var[kk*ntime] ;
				}
			}
			AFNI_OMP_END ;
#endif
			VECTIM_destroy(mrv) ;
			if(UNBP_OUT){ 
				DSET_write(outsetALL) ; if( verb ) WROTE_DSET(outsetALL) ;
			}
		}
	}// end of NumDen loop


	// @@
	INFO_message("Starting the (f)ALaFFel calcs") ;

	// allocations
	series1 = (double *)calloc(ntime,sizeof(double)); 
	series2 = (double *)calloc(ntime,sizeof(double)); 
	xx1 = (double *)calloc(2*ntime,sizeof(double)); 
	xx2 = (double *)calloc(2*ntime,sizeof(double)); 
	alff = (float *)calloc(nvox,sizeof(float)); 
	malff = (float *)calloc(nvox,sizeof(float)); 
	falff = (float *)calloc(nvox,sizeof(float)); 

	if( (series1 == NULL) || (series2 == NULL) 
		 || (xx1 == NULL) || (xx2 == NULL) 
		 || (alff == NULL) || (malff == NULL) || (falff == NULL)) { 
		fprintf(stderr, "\n\n MemAlloc failure.\n\n");
		exit(122);
	}
	if(DO_RSFA) {
		rsfa = (float *)calloc(nvox,sizeof(float)); 
		mrsfa = (float *)calloc(nvox,sizeof(float)); 
		frsfa = (float *)calloc(nvox,sizeof(float)); 
		if( (rsfa == NULL) || (mrsfa == NULL) || (frsfa == NULL)) { 
			fprintf(stderr, "\n\n MemAlloc failure.\n\n");
			exit(123);
		}	
	}
	
	
	work = gsl_fft_real_workspace_alloc (ntime);
	real1 = gsl_fft_real_wavetable_alloc (ntime);
	real2 = gsl_fft_real_wavetable_alloc (ntime);
	gsl_complex_packed_array compl_freqs1 = xx1;
	gsl_complex_packed_array compl_freqs2 = xx2;




	// *********************************************************************
	// *********************************************************************
	// **************    Falafelling = ALFF/fALFF calcs    *****************
	// *********************************************************************
	// *********************************************************************

	// Be now have the BP'ed data set (outset) and the non-BP'ed one
	// (outsetALL).  now we'll FFT both, get amplitudes in appropriate
	// ranges, and calculate:  ALFF, mALFF, fALFF,

	ctr = 0;
	for( kk=0; kk<nvox ; kk++) {
		if(mask[kk]) {
			
			// BP one, and unBP one, either for BP_LAST or !BP_LAST
			for( m=0 ; m<ntime ; m++ ) {
				series1[m] = THD_get_voxel(outset,kk,m);
				series2[m] = THD_get_voxel(outsetALL,kk,m);
			}
			
			
			mm = gsl_fft_real_transform(series1, 1, ntime, real1, work);
			mm = gsl_fft_halfcomplex_unpack(series1, compl_freqs1, 1, ntime);
			mm = gsl_fft_real_transform(series2, 1, ntime, real2, work);
			mm = gsl_fft_halfcomplex_unpack(series2, compl_freqs2, 1, ntime);

			numer = 0.0f; 
			denom = 0.0f;
			de_rsfa = 0.0f;
			nu_rsfa = 0.0f;
			for( m=1 ; m<N_ny ; m++ ) {
				mm = 2*m;
				pow2 = compl_freqs2[mm]*compl_freqs2[mm] +
					compl_freqs2[mm+1]*compl_freqs2[mm+1]; // power
				//pow2*=2;// factor of 2 since ampls are even funcs
				denom+= (float) sqrt(pow2); // amplitude 
				de_rsfa+= (float) pow2;
				
				if( ( m>=ind_low ) && ( m<=ind_high ) ){
					pow1 = compl_freqs1[mm]*compl_freqs1[mm]+
						compl_freqs1[mm+1]*compl_freqs1[mm+1];
					//pow1*=2;
					numer+= (float) sqrt(pow1);
					nu_rsfa+= (float) pow1;
				}
			}

			if( denom>0.000001 )
			  falff[kk] = numer/denom;
			else
			  falff[kk] = 0.;
			alff[kk] = 2*numer/sqnt;// factor of 2 since ampl is even funct
			meanALFF+= alff[kk];

			if(DO_RSFA){
			  nu_rsfa = sqrt(2*nu_rsfa); // factor of 2 since ampls 
			  de_rsfa = sqrt(2*de_rsfa); // are even funcs
			  if( de_rsfa>0.000001 )
			    frsfa[kk] = nu_rsfa/de_rsfa;
			  else
			    frsfa[kk]=0.;
			  rsfa[kk] = nu_rsfa/nt_fac;
			  meanRSFA+= rsfa[kk];
			}
			
			ctr+=1;
		}
	}
	meanALFF/= ctr;
	meanRSFA/= ctr;

	gsl_fft_real_wavetable_free(real1);
	gsl_fft_real_wavetable_free(real2);
	gsl_fft_real_workspace_free(work);

	// ALFFs divided by mean of brain value
	for( kk=0 ; kk<nvox ; kk++ ) 
		if(mask[kk]){
			malff[kk] = alff[kk]/meanALFF;
			if(DO_RSFA)
				mrsfa[kk] = rsfa[kk]/meanRSFA;
		}
	// **************************************************************
	// **************************************************************
	//                 Store and output
	// **************************************************************
	// **************************************************************
	
	outsetALFF = EDIT_empty_copy( inset ) ; 
	sprintf(out_alff,"%s_ALFF",prefix); 
	EDIT_dset_items( outsetALFF,
                    ADN_nvals, 1,
						  ADN_datum_all , MRI_float , 
						  ADN_prefix    , out_alff,
						  ADN_none ) ;
	if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetALFF)) )
		ERROR_exit("Can't overwrite existing dataset '%s'",
					  DSET_HEADNAME(outsetALFF));
	EDIT_substitute_brick(outsetALFF, 0, MRI_float, alff); 
	alff=NULL;
	THD_load_statistics(outsetALFF);
	tross_Make_History("3dRSFC", argc, argv, outsetALFF);
	THD_write_3dim_dataset(NULL, NULL, outsetALFF, True);

	outsetfALFF = EDIT_empty_copy( inset ) ;
	sprintf(out_falff,"%s_fALFF",prefix); 
	EDIT_dset_items( outsetfALFF,
                    ADN_nvals, 1,
						  ADN_datum_all , MRI_float , 
						  ADN_prefix    , out_falff,
						  ADN_none ) ;
	if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetfALFF)) )
		ERROR_exit("Can't overwrite existing dataset '%s'",
					  DSET_HEADNAME(outsetfALFF));
	EDIT_substitute_brick(outsetfALFF, 0, MRI_float, falff); 
	falff=NULL;
	THD_load_statistics(outsetfALFF);
	tross_Make_History("3dRSFC", argc, argv, outsetfALFF);
	THD_write_3dim_dataset(NULL, NULL, outsetfALFF, True);



	outsetmALFF = EDIT_empty_copy( inset ) ;
	sprintf(out_malff,"%s_mALFF",prefix); 
	EDIT_dset_items( outsetmALFF,
                    ADN_nvals, 1,
                    ADN_datum_all , MRI_float , 
						  ADN_prefix    , out_malff,
						  ADN_none ) ;
	if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetmALFF)) )
		ERROR_exit("Can't overwrite existing dataset '%s'",
					  DSET_HEADNAME(outsetmALFF));
	EDIT_substitute_brick(outsetmALFF, 0, MRI_float, malff); 
	malff=NULL;
	THD_load_statistics(outsetmALFF);
	tross_Make_History("3dRSFC", argc, argv, outsetmALFF);
	THD_write_3dim_dataset(NULL, NULL, outsetmALFF, True);

	if(DO_RSFA){
     outsetRSFA = EDIT_empty_copy( inset ) ;
		sprintf(out_rsfa,"%s_RSFA",prefix); 
		EDIT_dset_items( outsetRSFA,
                       ADN_nvals, 1,
                       ADN_datum_all , MRI_float , 
							  ADN_prefix    , out_rsfa,
							  ADN_none ) ;
		if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetRSFA)) )
			ERROR_exit("Can't overwrite existing dataset '%s'",
						  DSET_HEADNAME(outsetRSFA));
		EDIT_substitute_brick(outsetRSFA, 0, MRI_float, rsfa); 
		rsfa=NULL;
		THD_load_statistics(outsetRSFA);
		tross_Make_History("3dRSFC", argc, argv, outsetRSFA);
		THD_write_3dim_dataset(NULL, NULL, outsetRSFA, True);
		
      outsetfRSFA = EDIT_empty_copy( inset ) ;
		sprintf(out_frsfa,"%s_fRSFA",prefix); 
		EDIT_dset_items( outsetfRSFA,
                       ADN_nvals, 1,
                       ADN_datum_all , MRI_float , 
							  ADN_prefix    , out_frsfa,
							  ADN_none ) ;
		if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetfRSFA)) )
			ERROR_exit("Can't overwrite existing dataset '%s'",
						  DSET_HEADNAME(outsetfRSFA));
		EDIT_substitute_brick(outsetfRSFA, 0, MRI_float, frsfa); 
		frsfa=NULL;
		THD_load_statistics(outsetfRSFA);
		tross_Make_History("3dRSFC", argc, argv, outsetfRSFA);
		THD_write_3dim_dataset(NULL, NULL, outsetfRSFA, True);
		
		outsetmRSFA = EDIT_empty_copy( inset ) ; 
		sprintf(out_mrsfa,"%s_mRSFA",prefix); 
		EDIT_dset_items( outsetmRSFA,
                       ADN_nvals, 1,
                       ADN_datum_all , MRI_float , 
							  ADN_prefix    , out_mrsfa,
							  ADN_none ) ;
		if( !THD_ok_overwrite() && THD_is_ondisk(DSET_HEADNAME(outsetmRSFA)) )
			ERROR_exit("Can't overwrite existing dataset '%s'",
						  DSET_HEADNAME(outsetmRSFA));
		EDIT_substitute_brick(outsetmRSFA, 0, MRI_float, mrsfa); 
		mrsfa=NULL;
		THD_load_statistics(outsetmRSFA);
		tross_Make_History("3dRSFC", argc, argv, outsetmRSFA);
		THD_write_3dim_dataset(NULL, NULL, outsetmRSFA, True);
	}



	// ************************************************************
	// ************************************************************
	//                    Freeing
	// ************************************************************
	// ************************************************************

	DSET_delete(inset);
	DSET_delete(outsetALL);
	DSET_delete(outset);
	DSET_delete(outsetALFF);
	DSET_delete(outsetmALFF);
	DSET_delete(outsetfALFF);
	DSET_delete(outsetRSFA);
	DSET_delete(outsetmRSFA);
	DSET_delete(outsetfRSFA);

	free(inset);
	free(outsetALL);
	free(outset);
	free(outsetALFF);
	free(outsetmALFF);
	free(outsetfALFF);
	free(outsetRSFA);
	free(outsetmRSFA);
	free(outsetfRSFA);

	free(rsfa);
	free(mrsfa);
	free(frsfa);
	free(alff);
	free(malff);
	free(falff);
	free(mask);
	free(series1);
	free(series2);
	free(xx1);
	free(xx2);

	exit(0) ;
}
Пример #2
0
int main( int argc , char * argv[] )
{
   int do_norm=0 , qdet=2 , have_freq=0 , do_automask=0 ;
   float dt=0.0f , fbot=0.0f,ftop=999999.9f , blur=0.0f ;
   MRI_IMARR *ortar=NULL ; MRI_IMAGE *ortim=NULL ;
   THD_3dim_dataset **ortset=NULL ; int nortset=0 ;
   THD_3dim_dataset *inset=NULL , *outset ;
   char *prefix="bandpass" ;
   byte *mask=NULL ;
   int mask_nx=0,mask_ny=0,mask_nz=0,nmask , verb=1 , 
       nx,ny,nz,nvox , nfft=0 , kk ;
   float **vec , **ort=NULL ; int nort=0 , vv , nopt , ntime  ;
   MRI_vectim *mrv ;
   float pvrad=0.0f ; int nosat=0 ;
   int do_despike=0 ;

   /*-- help? --*/

   AFNI_SETUP_OMP(0) ;  /* 24 Jun 2013 */

   if( argc < 2 || strcmp(argv[1],"-help") == 0 ){
     printf(
       "\n"
       "** NOTA BENE:  For the purpose of preparing resting-state FMRI datasets **\n"
       "** for analysis (e.g., with 3dGroupInCorr),  this program is now mostly **\n"
       "** superseded by the afni_proc.py script.  See the 'afni_proc.py -help' **\n"
       "** section 'Resting state analysis (modern)' to get our current rs-FMRI **\n"
       "** pre-processing recommended sequence of steps. -- RW Cox, et alii.    **\n"
       "\n"
       "Usage: 3dBandpass [options] fbot ftop dataset\n"
       "\n"
       "* One function of this program is to prepare datasets for input\n"
       "   to 3dSetupGroupInCorr.  Other uses are left to your imagination.\n"
       "\n"
       "* 'dataset' is a 3D+time sequence of volumes\n"
       "   ++ This must be a single imaging run -- that is, no discontinuities\n"
       "       in time from 3dTcat-ing multiple datasets together.\n"
       "\n"
       "* fbot = lowest frequency in the passband, in Hz\n"
       "   ++ fbot can be 0 if you want to do a lowpass filter only;\n"
       "       HOWEVER, the mean and Nyquist freq are always removed.\n"
       "\n"
       "* ftop = highest frequency in the passband (must be > fbot)\n"
       "   ++ if ftop > Nyquist freq, then it's a highpass filter only.\n"
       "\n"
       "* Set fbot=0 and ftop=99999 to do an 'allpass' filter.\n"
       "  ++ Except for removal of the 0 and Nyquist frequencies, that is.\n"
       "\n"
       "* You cannot construct a 'notch' filter with this program!\n"
       "  ++ You could use 3dBandpass followed by 3dcalc to get the same effect.\n"
       "  ++ If you are understand what you are doing, that is.\n"
       "  ++ Of course, that is the AFNI way -- if you don't want to\n"
       "     understand what you are doing, use Some other PrograM, and\n"
       "     you can still get Fine StatisticaL maps.\n"
       "\n"
       "* 3dBandpass will fail if fbot and ftop are too close for comfort.\n"
       "  ++ Which means closer than one frequency grid step df,\n"
       "     where df = 1 / (nfft * dt) [of course]\n"
       "\n"
       "* The actual FFT length used will be printed, and may be larger\n"
       "   than the input time series length for the sake of efficiency.\n"
       "  ++ The program will use a power-of-2, possibly multiplied by\n"
       "     a power of 3 and/or 5 (up to and including the 3rd power of\n"
       "     each of these: 3, 9, 27, and 5, 25, 125).\n"
       "\n"
       "* Note that the results of combining 3dDetrend and 3dBandpass will\n"
       "   depend on the order in which you run these programs.  That's why\n"
       "   3dBandpass has the '-ort' and '-dsort' options, so that the\n"
       "   time series filtering can be done properly, in one place.\n"
       "\n"
       "* The output dataset is stored in float format.\n"
       "\n"
       "* The order of processing steps is the following (most are optional):\n"
       " (0) Check time series for initial transients [does not alter data]\n"
       " (1) Despiking of each time series\n"
       " (2) Removal of a constant+linear+quadratic trend in each time series\n"
       " (3) Bandpass of data time series\n"
       " (4) Bandpass of -ort time series, then detrending of data\n"
       "      with respect to the -ort time series\n"
       " (5) Bandpass and de-orting of the -dsort dataset,\n"
       "      then detrending of the data with respect to -dsort\n"
       " (6) Blurring inside the mask [might be slow]\n"
       " (7) Local PV calculation     [WILL be slow!]\n"
       " (8) L2 normalization         [will be fast.]\n"
       "\n"
       "--------\n"
       "OPTIONS:\n"
       "--------\n"
       " -despike        = Despike each time series before other processing.\n"
       "                   ++ Hopefully, you don't actually need to do this,\n"
       "                      which is why it is optional.\n"
       " -ort f.1D       = Also orthogonalize input to columns in f.1D\n"
       "                   ++ Multiple '-ort' options are allowed.\n"
       " -dsort fset     = Orthogonalize each voxel to the corresponding\n"
       "                    voxel time series in dataset 'fset', which must\n"
       "                    have the same spatial and temporal grid structure\n"
       "                    as the main input dataset.\n"
       "                   ++ At present, only one '-dsort' option is allowed.\n"
       " -nodetrend      = Skip the quadratic detrending of the input that\n"
       "                    occurs before the FFT-based bandpassing.\n"
       "                   ++ You would only want to do this if the dataset\n"
       "                      had been detrended already in some other program.\n"
       " -dt dd          = set time step to 'dd' sec [default=from dataset header]\n"
       " -nfft N         = set the FFT length to 'N' [must be a legal value]\n"
       " -norm           = Make all output time series have L2 norm = 1\n"
       "                   ++ i.e., sum of squares = 1\n"
       " -mask mset      = Mask dataset\n"
       " -automask       = Create a mask from the input dataset\n"
       " -blur fff       = Blur (inside the mask only) with a filter\n"
       "                    width (FWHM) of 'fff' millimeters.\n"
       " -localPV rrr    = Replace each vector by the local Principal Vector\n"
       "                    (AKA first singular vector) from a neighborhood\n"
       "                    of radius 'rrr' millimiters.\n"
       "                   ++ Note that the PV time series is L2 normalized.\n"
       "                   ++ This option is mostly for Bob Cox to have fun with.\n"
       "\n"
       " -input dataset  = Alternative way to specify input dataset.\n"
       " -band fbot ftop = Alternative way to specify passband frequencies.\n"
       "\n"
       " -prefix ppp     = Set prefix name of output dataset.\n"
       " -quiet          = Turn off the fun and informative messages. (Why?)\n"
       "\n"
       " -notrans        = Don't check for initial positive transients in the data:\n"
       "  *OR*             ++ The test is a little slow, so skipping it is OK,\n"
       " -nosat               if you KNOW the data time series are transient-free.\n"
       "                   ++ Or set AFNI_SKIP_SATCHECK to YES.\n"
       "                   ++ Initial transients won't be handled well by the\n"
       "                      bandpassing algorithm, and in addition may seriously\n"
       "                      contaminate any further processing, such as inter-voxel\n"
       "                      correlations via InstaCorr.\n"
       "                   ++ No other tests are made [yet] for non-stationary behavior\n"
       "                      in the time series data.\n"
     ) ;
     PRINT_AFNI_OMP_USAGE(
       "3dBandpass" ,
       "* At present, the only part of 3dBandpass that is parallelized is the\n"
       "  '-blur' option, which processes each sub-brick independently.\n"
     ) ;
     PRINT_COMPILE_DATE ; exit(0) ;
   }

   /*-- startup --*/

   mainENTRY("3dBandpass"); machdep();
   AFNI_logger("3dBandpass",argc,argv);
   PRINT_VERSION("3dBandpass"); AUTHOR("RW Cox");

   nosat =  AFNI_yesenv("AFNI_SKIP_SATCHECK") ;

   nopt = 1 ;
   while( nopt < argc && argv[nopt][0] == '-' ){

     if( strcmp(argv[nopt],"-despike") == 0 ){  /* 08 Oct 2010 */
       do_despike++ ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-nfft") == 0 ){
       int nnup ;
       if( ++nopt >= argc ) ERROR_exit("need an argument after -nfft!") ;
       nfft = (int)strtod(argv[nopt],NULL) ;
       nnup = csfft_nextup_even(nfft) ;
       if( nfft < 16 || nfft != nnup )
         ERROR_exit("value %d after -nfft is illegal! Next legal value = %d",nfft,nnup) ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-blur") == 0 ){
       if( ++nopt >= argc ) ERROR_exit("need an argument after -blur!") ;
       blur = strtod(argv[nopt],NULL) ;
       if( blur <= 0.0f ) WARNING_message("non-positive blur?!") ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-localPV") == 0 ){
       if( ++nopt >= argc ) ERROR_exit("need an argument after -localpv!") ;
       pvrad = strtod(argv[nopt],NULL) ;
       if( pvrad <= 0.0f ) WARNING_message("non-positive -localpv?!") ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-prefix") == 0 ){
       if( ++nopt >= argc ) ERROR_exit("need an argument after -prefix!") ;
       prefix = strdup(argv[nopt]) ;
       if( !THD_filename_ok(prefix) ) ERROR_exit("bad -prefix option!") ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-automask") == 0 ){
       if( mask != NULL ) ERROR_exit("Can't use -mask AND -automask!") ;
       do_automask = 1 ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-mask") == 0 ){
       THD_3dim_dataset *mset ;
       if( ++nopt >= argc ) ERROR_exit("Need argument after '-mask'") ;
       if( mask != NULL || do_automask ) ERROR_exit("Can't have two mask inputs") ;
       mset = THD_open_dataset( argv[nopt] ) ;
       CHECK_OPEN_ERROR(mset,argv[nopt]) ;
       DSET_load(mset) ; CHECK_LOAD_ERROR(mset) ;
       mask_nx = DSET_NX(mset); mask_ny = DSET_NY(mset); mask_nz = DSET_NZ(mset);
       mask = THD_makemask( mset , 0 , 0.5f, 0.0f ) ; DSET_delete(mset) ;
       if( mask == NULL ) ERROR_exit("Can't make mask from dataset '%s'",argv[nopt]) ;
       nmask = THD_countmask( mask_nx*mask_ny*mask_nz , mask ) ;
       if( verb ) INFO_message("Number of voxels in mask = %d",nmask) ;
       if( nmask < 1 ) ERROR_exit("Mask is too small to process") ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-norm") == 0 ){
       do_norm = 1 ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-quiet") == 0 ){
       verb = 0 ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-notrans") == 0 || strcmp(argv[nopt],"-nosat") == 0 ){
       nosat = 1 ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-ort") == 0 ){
       if( ++nopt >= argc ) ERROR_exit("need an argument after -ort!") ;
       if( ortar == NULL ) INIT_IMARR(ortar) ;
       ortim = mri_read_1D( argv[nopt] ) ;
       if( ortim == NULL ) ERROR_exit("can't read from -ort '%s'",argv[nopt]) ;
       mri_add_name(argv[nopt],ortim) ;
       ADDTO_IMARR(ortar,ortim) ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-dsort") == 0 ){
       THD_3dim_dataset *qset ;
       if( ++nopt >= argc ) ERROR_exit("need an argument after -dsort!") ;
       if( nortset > 0 ) ERROR_exit("only 1 -dsort option is allowed!") ;
       qset = THD_open_dataset(argv[nopt]) ;
       CHECK_OPEN_ERROR(qset,argv[nopt]) ;
       ortset = (THD_3dim_dataset **)realloc(ortset,
                                       sizeof(THD_3dim_dataset *)*(nortset+1)) ;
       ortset[nortset++] = qset ;
       nopt++ ; continue ;
     }

     if( strncmp(argv[nopt],"-nodetrend",6) == 0 ){
       qdet = 0 ; nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-dt") == 0 ){
       if( ++nopt >= argc ) ERROR_exit("need an argument after -dt!") ;
       dt = (float)strtod(argv[nopt],NULL) ;
       if( dt <= 0.0f ) WARNING_message("value after -dt illegal!") ;
       nopt++ ; continue ;
     }

     if( strcmp(argv[nopt],"-input") == 0 ){
       if( inset != NULL ) ERROR_exit("Can't have 2 -input options!") ;
       if( ++nopt >= argc ) ERROR_exit("need an argument after -input!") ;
       inset = THD_open_dataset(argv[nopt]) ;
       CHECK_OPEN_ERROR(inset,argv[nopt]) ;
       nopt++ ; continue ;
     }

     if( strncmp(argv[nopt],"-band",5) == 0 ){
       if( ++nopt >= argc-1 ) ERROR_exit("need 2 arguments after -band!") ;
       if( have_freq ) WARNING_message("second -band option replaces first one!") ;
       fbot = strtod(argv[nopt++],NULL) ;
       ftop = strtod(argv[nopt++],NULL) ;
       have_freq = 1 ; continue ;
     }

     ERROR_exit("Unknown option: '%s'",argv[nopt]) ;
   }

   /** check inputs for reasonablositiness **/

   if( !have_freq ){
     if( nopt+1 >= argc )
       ERROR_exit("Need frequencies on command line after options!") ;
     fbot = (float)strtod(argv[nopt++],NULL) ;
     ftop = (float)strtod(argv[nopt++],NULL) ;
   }

   if( inset == NULL ){
     if( nopt >= argc )
       ERROR_exit("Need input dataset name on command line after options!") ;
     inset = THD_open_dataset(argv[nopt]) ;
     CHECK_OPEN_ERROR(inset,argv[nopt]) ; nopt++ ;
   }
   DSET_UNMSEC(inset) ;

   if( fbot < 0.0f  ) ERROR_exit("fbot value can't be negative!") ;
   if( ftop <= fbot ) ERROR_exit("ftop value %g must be greater than fbot value %g!",ftop,fbot) ;

   ntime = DSET_NVALS(inset) ;
   if( ntime < 9 ) ERROR_exit("Input dataset is too short!") ;

   if( nfft <= 0 ){
     nfft = csfft_nextup_even(ntime) ;
     if( verb ) INFO_message("Data length = %d  FFT length = %d",ntime,nfft) ;
     (void)THD_bandpass_set_nfft(nfft) ;
   } else if( nfft < ntime ){
     ERROR_exit("-nfft %d is less than data length = %d",nfft,ntime) ;
   } else {
     kk = THD_bandpass_set_nfft(nfft) ;
     if( kk != nfft && verb )
       INFO_message("Data length = %d  FFT length = %d",ntime,kk) ;
   }

   if( dt <= 0.0f ){
     dt = DSET_TR(inset) ;
     if( dt <= 0.0f ){
       WARNING_message("Setting dt=1.0 since input dataset lacks a time axis!") ;
       dt = 1.0f ;
     }
   }

   if( !THD_bandpass_OK(ntime,dt,fbot,ftop,1) ) ERROR_exit("Can't continue!") ;

   nx = DSET_NX(inset); ny = DSET_NY(inset); nz = DSET_NZ(inset); nvox = nx*ny*nz;

   /* check mask, or create it */

   if( verb ) INFO_message("Loading input dataset time series" ) ;
   DSET_load(inset) ;

   if( mask != NULL ){
     if( mask_nx != nx || mask_ny != ny || mask_nz != nz )
       ERROR_exit("-mask dataset grid doesn't match input dataset") ;

   } else if( do_automask ){
     mask = THD_automask( inset ) ;
     if( mask == NULL )
       ERROR_message("Can't create -automask from input dataset?") ;
     nmask = THD_countmask( DSET_NVOX(inset) , mask ) ;
     if( verb ) INFO_message("Number of voxels in automask = %d",nmask);
     if( nmask < 1 ) ERROR_exit("Automask is too small to process") ;

   } else {
     mask = (byte *)malloc(sizeof(byte)*nvox) ; nmask = nvox ;
     memset(mask,1,sizeof(byte)*nvox) ;
     if( verb ) INFO_message("No mask ==> processing all %d voxels",nvox);
   }

   /* A simple check of dataset quality [08 Feb 2010] */

   if( !nosat ){
     float val ;
     INFO_message(
      "Checking dataset for initial transients [use '-notrans' to skip this test]") ;
     val = THD_saturation_check(inset,mask,0,0) ; kk = (int)(val+0.54321f) ;
     if( kk > 0 )
       ININFO_message(
        "Looks like there %s %d non-steady-state initial time point%s :-(" ,
        ((kk==1) ? "is" : "are") , kk , ((kk==1) ? " " : "s") ) ;
     else if( val > 0.3210f )  /* don't ask where this threshold comes from! */
       ININFO_message(
        "MAYBE there's an initial positive transient of 1 point, but it's hard to tell\n") ;
     else
       ININFO_message("No widespread initial positive transient detected :-)") ;
   }

   /* check -dsort inputs for match to inset */

   for( kk=0 ; kk < nortset ; kk++ ){
     if( DSET_NX(ortset[kk])    != nx ||
         DSET_NY(ortset[kk])    != ny ||
         DSET_NZ(ortset[kk])    != nz ||
         DSET_NVALS(ortset[kk]) != ntime )
       ERROR_exit("-dsort %s doesn't match input dataset grid" ,
                  DSET_BRIKNAME(ortset[kk]) ) ;
   }

   /* convert input dataset to a vectim, which is more fun */

   mrv = THD_dset_to_vectim( inset , mask , 0 ) ;
   if( mrv == NULL ) ERROR_exit("Can't load time series data!?") ;
   DSET_unload(inset) ;

   /* similarly for the ort vectors */

   if( ortar != NULL ){
     for( kk=0 ; kk < IMARR_COUNT(ortar) ; kk++ ){
       ortim = IMARR_SUBIM(ortar,kk) ;
       if( ortim->nx < ntime )
         ERROR_exit("-ort file %s is shorter than input dataset time series",
                    ortim->name ) ;
       ort  = (float **)realloc( ort , sizeof(float *)*(nort+ortim->ny) ) ;
       for( vv=0 ; vv < ortim->ny ; vv++ )
         ort[nort++] = MRI_FLOAT_PTR(ortim) + ortim->nx * vv ;
     }
   }

   /* check whether processing leaves any DoF remaining  18 Mar 2015 [rickr] */
   {
      int nbprem = THD_bandpass_remain_dim(ntime, dt, fbot, ftop, 1);
      int bpused, nremain;
      int wlimit;               /* warning limit */

      bpused = ntime - nbprem;  /* #dim lost in bandpass step */

      nremain = nbprem - nort;  /* #dim left in output */
      if( nortset == 1 ) nremain--;
      nremain -= (qdet+1);

      if( verb ) INFO_message("%d dimensional data reduced to %d by:\n"
                    "    %d (bandpass), %d (-ort), %d (-dsort), %d (detrend)",
                    ntime, nremain, bpused, nort, nortset?1:0, qdet+1);

      /* possibly warn (if 95% lost) user or fail */
      wlimit = ntime/20;
      if( wlimit < 3 ) wlimit = 3;
      if( nremain < wlimit && nremain > 0 )
         WARNING_message("dimensionality reduced from %d to %d, be careful!",
                         ntime, nremain);
      if( nremain <= 0 ) /* FAILURE */
         ERROR_exit("dimensionality reduced from %d to %d, failing!",
                    ntime, nremain);
   }

   /* all the real work now */

   if( do_despike ){
     int_pair nsp ;
     if( verb ) INFO_message("Testing data time series for spikes") ;
     nsp = THD_vectim_despike9( mrv ) ;
     if( verb ) ININFO_message(" -- Squashed %d spikes from %d voxels",nsp.j,nsp.i) ;
   }

   if( verb ) INFO_message("Bandpassing data time series") ;
   (void)THD_bandpass_vectim( mrv , dt,fbot,ftop , qdet , nort,ort ) ;

   /* OK, maybe a little more work */

   if( nortset == 1 ){
     MRI_vectim *orv ;
     orv = THD_dset_to_vectim( ortset[0] , mask , 0 ) ;
     if( orv == NULL ){
       ERROR_message("Can't load -dsort %s",DSET_BRIKNAME(ortset[0])) ;
     } else {
       float *dp , *mvv , *ovv , ff ;
       if( verb ) INFO_message("Orthogonalizing to bandpassed -dsort") ;
       (void)THD_bandpass_vectim( orv , dt,fbot,ftop , qdet , nort,ort ) ;
       THD_vectim_normalize( orv ) ;
       dp = malloc(sizeof(float)*mrv->nvec) ;
       THD_vectim_vectim_dot( mrv , orv , dp ) ;
       for( vv=0 ; vv < mrv->nvec ; vv++ ){
         ff = dp[vv] ;
         if( ff != 0.0f ){
           mvv = VECTIM_PTR(mrv,vv) ; ovv = VECTIM_PTR(orv,vv) ;
           for( kk=0 ; kk < ntime ; kk++ ) mvv[kk] -= ff*ovv[kk] ;
         }
       }
       VECTIM_destroy(orv) ; free(dp) ;
     }
   }

   if( blur > 0.0f ){
     if( verb )
       INFO_message("Blurring time series data spatially; FWHM=%.2f",blur) ;
     mri_blur3D_vectim( mrv , blur ) ;
   }
   if( pvrad > 0.0f ){
     if( verb )
       INFO_message("Local PV-ing time series data spatially; radius=%.2f",pvrad) ;
     THD_vectim_normalize( mrv ) ;
     THD_vectim_localpv( mrv , pvrad ) ;
   }
   if( do_norm && pvrad <= 0.0f ){
     if( verb ) INFO_message("L2 normalizing time series data") ;
     THD_vectim_normalize( mrv ) ;
   }

   /* create output dataset, populate it, write it, then quit */

   if( verb ) INFO_message("Creating output dataset in memory, then writing it") ;
   outset = EDIT_empty_copy(inset) ;
   /* do not copy scalars    11 Sep 2015 [rickr] */
   EDIT_dset_items( outset , ADN_prefix,prefix ,
                             ADN_brick_fac,NULL ,
                    ADN_none ) ;
   tross_Copy_History( inset , outset ) ;
   tross_Make_History( "3dBandpass" , argc,argv , outset ) ;

   for( vv=0 ; vv < ntime ; vv++ )
     EDIT_substitute_brick( outset , vv , MRI_float , NULL ) ;

#if 1
   THD_vectim_to_dset( mrv , outset ) ;
#else
 AFNI_OMP_START ;
#pragma omp parallel
 { float *far , *var ; int *ivec=mrv->ivec ; int vv,kk ;
#pragma omp for
   for( vv=0 ; vv < ntime ; vv++ ){
     far = DSET_BRICK_ARRAY(outset,vv) ; var = mrv->fvec + vv ;
     for( kk=0 ; kk < nmask ; kk++ ) far[ivec[kk]] = var[kk*ntime] ;
   }
 }
 AFNI_OMP_END ;
#endif
   VECTIM_destroy(mrv) ;
   DSET_write(outset) ; if( verb ) WROTE_DSET(outset) ;

   exit(0) ;
}