int
tend_msimMain(int argc, const char **argv, const char *me,
hestParm *hparm) {
int pret;
hestOpt *hopt = NULL;
char *perr, *err;
airArray *mop;
tenExperSpec *espec;
const tenModel *model;
int E, seed, keyValueSet, outType, plusB0, insertB0;
Nrrd *nin, *nT2, *_ngrad, *ngrad, *nout;
char *outS, *modS;
double bval, sigma;
/* maybe this can go in tend.c, but for some reason its explicitly
set to AIR_FALSE there */
hparm->elideSingleOtherDefault = AIR_TRUE;
hestOptAdd(&hopt, "sigma", "sigma", airTypeDouble, 1, 1, &sigma, "0.0",
"Gaussian/Rician noise parameter");
hestOptAdd(&hopt, "seed", "seed", airTypeInt, 1, 1, &seed, "42",
"seed value for RNG which creates noise");
hestOptAdd(&hopt, "g", "grad list", airTypeOther, 1, 1, &_ngrad, NULL,
"gradient list, one row per diffusion-weighted image",
NULL, NULL, nrrdHestNrrd);
hestOptAdd(&hopt, "b0", "b0 image", airTypeOther, 1, 1, &nT2, "",
"reference non-diffusion-weighted (\"B0\") image, which "
"may be needed if it isn't part of give model param image",
NULL, NULL, nrrdHestNrrd);
hestOptAdd(&hopt, "i", "model image", airTypeOther, 1, 1, &nin, "-",
"input model image", NULL, NULL, nrrdHestNrrd);
hestOptAdd(&hopt, "m", "model", airTypeString, 1, 1, &modS, NULL,
"model with which to simulate DWIs, which must be specified if "
"it is not indicated by the first axis in input model image.");
hestOptAdd(&hopt, "ib0", "bool", airTypeBool, 1, 1, &insertB0, "false",
"insert a non-DW B0 image at the beginning of the experiment "
"specification (useful if the given gradient list doesn't "
"already have one) and hence also insert a B0 image at the "
"beginning of the output simulated DWIs");
hestOptAdd(&hopt, "b", "b", airTypeDouble, 1, 1, &bval, "1000",
"b value for simulated scan");
hestOptAdd(&hopt, "kvp", "bool", airTypeBool, 1, 1, &keyValueSet, "true",
"generate key/value pairs in the NRRD header corresponding "
"to the input b-value and gradients.");
hestOptAdd(&hopt, "t", "type", airTypeEnum, 1, 1, &outType, "float",
"output type of DWIs", NULL, nrrdType);
hestOptAdd(&hopt, "o", "nout", airTypeString, 1, 1, &outS, "-",
"output dwis");
mop = airMopNew();
airMopAdd(mop, hopt, (airMopper)hestOptFree, airMopAlways);
USAGE(_tend_msimInfoL);
PARSE();
airMopAdd(mop, hopt, (airMopper)hestParseFree, airMopAlways);
nout = nrrdNew();
airMopAdd(mop, nout, (airMopper)nrrdNuke, airMopAlways);
espec = tenExperSpecNew();
airMopAdd(mop, espec, (airMopper)tenExperSpecNix, airMopAlways);
airSrandMT(seed);
if (nrrdTypeDouble == _ngrad->type) {
ngrad = _ngrad;
} else {
ngrad = nrrdNew();
airMopAdd(mop, ngrad, (airMopper)nrrdNuke, airMopAlways);
if (nrrdConvert(ngrad, _ngrad, nrrdTypeDouble)) {
airMopAdd(mop, err=biffGetDone(NRRD), airFree, airMopAlways);
fprintf(stderr, "%s: trouble converting grads to %s:\n%s\n", me,
airEnumStr(nrrdType, nrrdTypeDouble), err);
airMopError(mop); return 1;
}
}
plusB0 = AIR_FALSE;
if (airStrlen(modS)) {
if (tenModelParse(&model, &plusB0, AIR_FALSE, modS)) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble parsing model \"%s\":\n%s\n",
me, modS, err);
airMopError(mop); return 1;
}
} else if (tenModelFromAxisLearnPossible(nin->axis + 0)) {
if (tenModelFromAxisLearn(&model, &plusB0, nin->axis + 0)) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble parsing model frmo axis 0 of nin:\n%s\n",
me, err);
airMopError(mop); return 1;
}
} else {
fprintf(stderr, "%s: need model specified either via \"-m\" or input "
"model image axis 0\n", me);
airMopError(mop); return 1;
}
/* we have learned plusB0, but we don't actually need it;
either: it describes the given model param image
(which is courteous but not necessary since the logic inside
tenModeSimulate will see this),
or: it is trying to say something about including B0 amongst
model parameters (which isn't actually meaningful in the
context of simulated DWIs */
E = 0;
if (!E) E |= tenGradientCheck(ngrad, nrrdTypeDouble, 1);
if (!E) E |= tenExperSpecGradSingleBValSet(espec, insertB0, bval,
AIR_CAST(const double *,
ngrad->data),
ngrad->axis[1].size);
if (!E) E |= tenModelSimulate(nout, outType, espec,
model, nT2, nin, keyValueSet);
if (E) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble:\n%s\n", me, err);
airMopError(mop); return 1;
}
if (nrrdSave(outS, nout, NULL)) {
airMopAdd(mop, err=biffGetDone(NRRD), airFree, airMopAlways);
fprintf(stderr, "%s: trouble writing:\n%s\n", me, err);
airMopError(mop); return 1;
}
airMopOkay(mop);
return 0;
}
int
tend_mfitMain(int argc, char **argv, char *me, hestParm *hparm) {
int pret;
hestOpt *hopt = NULL;
char *perr, *err;
airArray *mop;
Nrrd *nin, *nout, *nterr;
char *outS, *terrS, *modS;
int knownB0, saveB0, verbose, mlfit, typeOut;
unsigned int maxIter, minIter, starts;
double sigma, eps;
const tenModel *model;
tenExperSpec *espec;
hestOptAdd(&hopt, "v", "verbose", airTypeInt, 1, 1, &verbose, "0",
"verbosity level");
hestOptAdd(&hopt, "m", "model", airTypeString, 1, 1, &modS, NULL,
"which model to fit. Use optional \"b0+\" prefix to "
"indicate that the B0 image should also be saved.");
hestOptAdd(&hopt, "ns", "# starts", airTypeUInt, 1, 1, &starts, "1",
"number of random starting points at which to initialize "
"fitting");
hestOptAdd(&hopt, "ml", NULL, airTypeInt, 0, 0, &mlfit, NULL,
"do ML fitting, rather than least-squares, which also "
"requires setting \"-sigma\"");
hestOptAdd(&hopt, "sigma", "sigma", airTypeDouble, 1, 1, &sigma, "nan",
"Rician noise parameter");
hestOptAdd(&hopt, "eps", "eps", airTypeDouble, 1, 1, &eps, "0.01",
"convergence epsilon");
hestOptAdd(&hopt, "mini", "min iters", airTypeUInt, 1, 1, &minIter, "3",
"minimum required # iterations for fitting.");
hestOptAdd(&hopt, "maxi", "max iters", airTypeUInt, 1, 1, &maxIter, "100",
"maximum allowable # iterations for fitting.");
hestOptAdd(&hopt, "knownB0", "bool", airTypeBool, 1, 1, &knownB0, NULL,
"Indicates if the B=0 non-diffusion-weighted reference image "
"is known (\"true\"), or if it has to be estimated along with "
"the other model parameters (\"false\")");
hestOptAdd(&hopt, "t", "type", airTypeEnum, 1, 1, &typeOut, "float",
"output type of model parameters",
NULL, nrrdType);
hestOptAdd(&hopt, "i", "dwi", airTypeOther, 1, 1, &nin, "-",
"all the diffusion-weighted images in one 4D nrrd",
NULL, NULL, nrrdHestNrrd);
hestOptAdd(&hopt, "o", "nout", airTypeString, 1, 1, &outS, "-",
"output tensor volume");
hestOptAdd(&hopt, "eo", "filename", airTypeString, 1, 1, &terrS, "",
"Giving a filename here allows you to save out the per-sample "
"fitting error. By default, no such error is saved.");
mop = airMopNew();
airMopAdd(mop, hopt, (airMopper)hestOptFree, airMopAlways);
USAGE(_tend_mfitInfoL);
JUSTPARSE();
airMopAdd(mop, hopt, (airMopper)hestParseFree, airMopAlways);
nterr = NULL;
espec = tenExperSpecNew();
airMopAdd(mop, espec, (airMopper)tenExperSpecNix, airMopAlways);
nout = nrrdNew();
airMopAdd(mop, nout, (airMopper)nrrdNuke, airMopAlways);
if (tenModelParse(&model, &saveB0, AIR_FALSE, modS)) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble parsing model \"%s\":\n%s\n", me, modS, err);
airMopError(mop); return 1;
}
if (tenExperSpecFromKeyValueSet(espec, nin)) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble getting exper from kvp:\n%s\n", me, err);
airMopError(mop); return 1;
}
if (tenModelSqeFit(nout,
airStrlen(terrS) ? &nterr : NULL,
model, espec, nin,
knownB0, saveB0, typeOut,
minIter, maxIter, starts, eps, NULL)) {
airMopAdd(mop, err=biffGetDone(TEN), airFree, airMopAlways);
fprintf(stderr, "%s: trouble fitting:\n%s\n", me, err);
airMopError(mop); return 1;
}
if (nrrdSave(outS, nout, NULL)
|| (airStrlen(terrS) && nrrdSave(terrS, nterr, NULL))) {
airMopAdd(mop, err=biffGetDone(NRRD), airFree, airMopAlways);
fprintf(stderr, "%s: trouble writing output:\n%s\n", me, err);
airMopError(mop); return 1;
}
airMopOkay(mop);
return 0;
}
