| File: | src/ten/chan.c |
| Location: | line 155, column 3 |
| Description: | Value stored to 'val' is never read |
| 1 | /* |
| 2 | Teem: Tools to process and visualize scientific data and images . |
| 3 | Copyright (C) 2013, 2012, 2011, 2010, 2009 University of Chicago |
| 4 | Copyright (C) 2008, 2007, 2006, 2005 Gordon Kindlmann |
| 5 | Copyright (C) 2004, 2003, 2002, 2001, 2000, 1999, 1998 University of Utah |
| 6 | |
| 7 | This library is free software; you can redistribute it and/or |
| 8 | modify it under the terms of the GNU Lesser General Public License |
| 9 | (LGPL) as published by the Free Software Foundation; either |
| 10 | version 2.1 of the License, or (at your option) any later version. |
| 11 | The terms of redistributing and/or modifying this software also |
| 12 | include exceptions to the LGPL that facilitate static linking. |
| 13 | |
| 14 | This library is distributed in the hope that it will be useful, |
| 15 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 17 | Lesser General Public License for more details. |
| 18 | |
| 19 | You should have received a copy of the GNU Lesser General Public License |
| 20 | along with this library; if not, write to Free Software Foundation, Inc., |
| 21 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 22 | */ |
| 23 | |
| 24 | #include "ten.h" |
| 25 | #include "privateTen.h" |
| 26 | |
| 27 | const char * |
| 28 | tenDWMRIModalityKey = "modality"; |
| 29 | |
| 30 | const char * |
| 31 | tenDWMRIModalityVal = "DWMRI"; |
| 32 | |
| 33 | const char * |
| 34 | tenDWMRINAVal = "n/a"; |
| 35 | |
| 36 | const char * |
| 37 | tenDWMRIBValueKey = "DWMRI_b-value"; |
| 38 | |
| 39 | const char * |
| 40 | tenDWMRIGradKeyFmt = "DWMRI_gradient_%04u"; |
| 41 | |
| 42 | const char * |
| 43 | tenDWMRIBmatKeyFmt = "DWMRI_B-matrix_%04u"; |
| 44 | |
| 45 | const char * |
| 46 | tenDWMRINexKeyFmt = "DWMRI_NEX_%04u"; |
| 47 | |
| 48 | const char * |
| 49 | tenDWMRISkipKeyFmt = "DWMRI_skip_%04u"; |
| 50 | |
| 51 | /* |
| 52 | ******** tenDWMRIKeyValueParse |
| 53 | ** |
| 54 | ** Parses through key-value pairs in the NRRD header to determine the |
| 55 | ** list of diffusion-sensitizing gradient directions, or B-matrices |
| 56 | ** (depending to what was found), according the NAMIC conventions. |
| 57 | ** This requires, among other things, that ndwi be have exactly one |
| 58 | ** axis with kind nrrdKindList (or nrrdKindVector), which is taken to |
| 59 | ** be the DWI axis. |
| 60 | ** |
| 61 | ** Either *ngradP or *nbmatP is set to a newly- allocated nrrd |
| 62 | ** containing this information, and the other one is set to NULL |
| 63 | ** The (scalar) b-value is stored in *bP. The image values that are |
| 64 | ** to be skipped are stored in the *skipP array (allocated here), |
| 65 | ** the length of that array is stored in *skipNumP. Unlike the skip |
| 66 | ** array used internally with tenEstimate, this is just a simple 1-D |
| 67 | ** array; it is not a list of pairs of (index,skipBool). |
| 68 | */ |
| 69 | int |
| 70 | tenDWMRIKeyValueParse(Nrrd **ngradP, Nrrd **nbmatP, double *bP, |
| 71 | unsigned int **skipP, unsigned int *skipNumP, |
| 72 | const Nrrd *ndwi) { |
| 73 | static const char me[]="tenDWMRIKeyValueParse"; |
| 74 | char tmpKey[AIR_STRLEN_MED(256+1)], |
| 75 | key[AIR_STRLEN_MED(256+1)], *val; |
| 76 | const char *keyFmt; |
| 77 | int dwiAxis; |
| 78 | unsigned int axi, dwiIdx, dwiNum, valNum, valIdx, parsedNum, |
| 79 | nexNum, nexIdx, skipIdx, *skipLut; |
| 80 | Nrrd *ninfo; |
| 81 | double *info, normMax, norm; |
| 82 | airArray *mop, *skipArr; |
| 83 | |
| 84 | if (!( ngradP && nbmatP && skipP && skipNumP && bP && ndwi )) { |
| 85 | biffAddf(TENtenBiffKey, "%s: got NULL pointer", me); |
| 86 | return 1; |
| 87 | } |
| 88 | |
| 89 | /* check modality */ |
| 90 | val = nrrdKeyValueGet(ndwi, tenDWMRIModalityKey); |
| 91 | if (!val) { |
| 92 | biffAddf(TENtenBiffKey, "%s: didn't have \"%s\" key", me, tenDWMRIModalityKey); |
| 93 | return 1; |
| 94 | } |
| 95 | if (strncmp(tenDWMRIModalityVal, val + strspn(val, AIR_WHITESPACE" \t\n\r\v\f"), |
| 96 | strlen(tenDWMRIModalityVal))) { |
| 97 | biffAddf(TENtenBiffKey, "%s: \"%s\" value was \"%s\", not \"%s\"", me, |
| 98 | tenDWMRIModalityKey, val, tenDWMRIModalityVal); |
| 99 | return 1; |
| 100 | } |
| 101 | val = (char *)airFree(val); |
| 102 | |
| 103 | /* learn b-value */ |
| 104 | val = nrrdKeyValueGet(ndwi, tenDWMRIBValueKey); |
| 105 | if (!val) { |
| 106 | biffAddf(TENtenBiffKey, "%s: didn't have \"%s\" key", me, tenDWMRIBValueKey); |
| 107 | return 1; |
| 108 | } |
| 109 | if (1 != sscanf(val, "%lg", bP)) { |
| 110 | biffAddf(TENtenBiffKey, "%s: couldn't parse float from value \"%s\" " |
| 111 | "for key \"%s\"", me, |
| 112 | val, tenDWMRIBValueKey); |
| 113 | return 1; |
| 114 | } |
| 115 | val = (char *)airFree(val); |
| 116 | |
| 117 | /* find single DWI axis, set dwiNum to its size */ |
| 118 | dwiAxis = -1; |
| 119 | for (axi=0; axi<ndwi->dim; axi++) { |
| 120 | /* the use of nrrdKindVector here is out of deference to how ITK's |
| 121 | itkNrrdImageIO.cxx uses nrrdKindVector for VECTOR pixels */ |
| 122 | if (nrrdKindList == ndwi->axis[axi].kind |
| 123 | || nrrdKindVector == ndwi->axis[axi].kind) { |
| 124 | if (-1 != dwiAxis) { |
| 125 | biffAddf(TENtenBiffKey, "%s: already saw %s or %s kind on axis %d", me, |
| 126 | airEnumStr(nrrdKind, nrrdKindList), |
| 127 | airEnumStr(nrrdKind, nrrdKindVector), dwiAxis); |
| 128 | return 1; |
| 129 | } |
| 130 | dwiAxis = axi; |
| 131 | } |
| 132 | } |
| 133 | if (-1 == dwiAxis) { |
| 134 | biffAddf(TENtenBiffKey, "%s: did not see \"%s\" kind on any axis", me, |
| 135 | airEnumStr(nrrdKind, nrrdKindList)); |
| 136 | return 1; |
| 137 | } |
| 138 | dwiNum = ndwi->axis[dwiAxis].size; |
| 139 | |
| 140 | /* figure out if we're parsing gradients or b-matrices */ |
| 141 | sprintf(tmpKey, tenDWMRIGradKeyFmt, 0)__builtin___sprintf_chk (tmpKey, 0, __builtin_object_size (tmpKey , 2 > 1 ? 1 : 0), tenDWMRIGradKeyFmt, 0); |
| 142 | val = nrrdKeyValueGet(ndwi, tmpKey); |
| 143 | if (val) { |
| 144 | valNum = 3; |
| 145 | } else { |
| 146 | valNum = 6; |
| 147 | sprintf(key, tenDWMRIBmatKeyFmt, 0)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), tenDWMRIBmatKeyFmt, 0); |
| 148 | val = nrrdKeyValueGet(ndwi, key); |
| 149 | if (!val) { |
| 150 | biffAddf(TENtenBiffKey, "%s: saw neither \"%s\" nor \"%s\" key", me, |
| 151 | tmpKey, key); |
| 152 | return 1; |
| 153 | } |
| 154 | } |
| 155 | val = (char *)airFree(val); |
Value stored to 'val' is never read | |
| 156 | |
| 157 | /* set up parsing and allocate one of output nrrds */ |
| 158 | if (3 == valNum) { |
| 159 | keyFmt = tenDWMRIGradKeyFmt; |
| 160 | ninfo = *ngradP = nrrdNew(); |
| 161 | *nbmatP = NULL((void*)0); |
| 162 | } else { |
| 163 | keyFmt = tenDWMRIBmatKeyFmt; |
| 164 | *ngradP = NULL((void*)0); |
| 165 | ninfo = *nbmatP = nrrdNew(); |
| 166 | } |
| 167 | if (nrrdMaybeAlloc_va(ninfo, nrrdTypeDouble, 2, |
| 168 | AIR_CAST(size_t, valNum)((size_t)(valNum)), |
| 169 | AIR_CAST(size_t, dwiNum)((size_t)(dwiNum)))) { |
| 170 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't allocate output", me); |
| 171 | return 1; |
| 172 | } |
| 173 | info = (double *)(ninfo->data); |
| 174 | |
| 175 | /* set up skip list recording */ |
| 176 | mop = airMopNew(); |
| 177 | skipArr = airArrayNew((void**)skipP, skipNumP, sizeof(unsigned int), 16); |
| 178 | airMopAdd(mop, skipArr, (airMopper)airArrayNix, airMopAlways); |
| 179 | skipLut = AIR_CALLOC(dwiNum, unsigned int)(unsigned int*)(calloc((dwiNum), sizeof(unsigned int))); |
| 180 | airMopAdd(mop, skipLut, airFree, airMopAlways); |
| 181 | if (!skipLut) { |
| 182 | biffAddf(TENtenBiffKey, "%s: couldn't allocate skip LUT", me); |
| 183 | airMopError(mop); return 1; |
| 184 | } |
| 185 | |
| 186 | /* parse values in ninfo */ |
| 187 | for (dwiIdx=0; dwiIdx<dwiNum; dwiIdx++) { |
| 188 | sprintf(key, keyFmt, dwiIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), keyFmt, dwiIdx); |
| 189 | val = nrrdKeyValueGet(ndwi, key); |
| 190 | if (!val) { |
| 191 | biffAddf(TENtenBiffKey, "%s: didn't see \"%s\" key", me, key); |
| 192 | airMopError(mop); return 1; |
| 193 | } |
| 194 | airToLower(val); |
| 195 | if (!strncmp(tenDWMRINAVal, val + strspn(val, AIR_WHITESPACE" \t\n\r\v\f"), |
| 196 | strlen(tenDWMRINAVal))) { |
| 197 | /* have no sensible gradient or B-matrix info here, and must skip */ |
| 198 | for (valIdx=0; valIdx<valNum; valIdx++) { |
| 199 | info[valIdx] = AIR_NAN(airFloatQNaN.f); |
| 200 | } |
| 201 | skipIdx = airArrayLenIncr(skipArr, 1); |
| 202 | (*skipP)[skipIdx] = dwiIdx; |
| 203 | skipLut[dwiIdx] = AIR_TRUE1; |
| 204 | /* can't have NEX on a skipped gradient or B-matrix */ |
| 205 | val = (char *)airFree(val); |
| 206 | sprintf(key, tenDWMRINexKeyFmt, dwiIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), tenDWMRINexKeyFmt, dwiIdx); |
| 207 | val = nrrdKeyValueGet(ndwi, key); |
| 208 | if (val) { |
| 209 | biffAddf(TENtenBiffKey, "%s: can't have NEX of skipped DWI %u", me, skipIdx); |
| 210 | airMopError(mop); return 1; |
| 211 | } |
| 212 | nexNum = 1; /* for "info +=" below */ |
| 213 | } else { |
| 214 | /* we probably do have sensible gradient or B-matrix info */ |
| 215 | parsedNum = airParseStrD(info, val, AIR_WHITESPACE" \t\n\r\v\f", valNum); |
| 216 | if (valNum != parsedNum) { |
| 217 | biffAddf(TENtenBiffKey, "%s: couldn't parse %d floats in value \"%s\" " |
| 218 | "for key \"%s\" (only got %d)", |
| 219 | me, valNum, val, key, parsedNum); |
| 220 | airMopError(mop); return 1; |
| 221 | } |
| 222 | val = (char *)airFree(val); |
| 223 | sprintf(key, tenDWMRINexKeyFmt, dwiIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), tenDWMRINexKeyFmt, dwiIdx); |
| 224 | val = nrrdKeyValueGet(ndwi, key); |
| 225 | if (!val) { |
| 226 | /* there is no NEX indicated */ |
| 227 | nexNum = 1; |
| 228 | } else { |
| 229 | if (1 != sscanf(val, "%u", &nexNum)) { |
| 230 | biffAddf(TENtenBiffKey, "%s: couldn't parse integer in value \"%s\" " |
| 231 | "for key \"%s\"", me, val, key); |
| 232 | airMopError(mop); return 1; |
| 233 | } |
| 234 | val = (char *)airFree(val); |
| 235 | if (!( nexNum >= 1 )) { |
| 236 | biffAddf(TENtenBiffKey, "%s: NEX (%d) for DWI %d not >= 1", |
| 237 | me, nexNum, dwiIdx); |
| 238 | airMopError(mop); return 1; |
| 239 | } |
| 240 | if (!( dwiIdx + nexNum - 1 < dwiNum )) { |
| 241 | biffAddf(TENtenBiffKey, "%s: NEX %d for DWI %d implies %d DWI > real # DWI %d", |
| 242 | me, nexNum, dwiIdx, dwiIdx + nexNum, dwiNum); |
| 243 | airMopError(mop); return 1; |
| 244 | } |
| 245 | for (nexIdx=1; nexIdx<nexNum; nexIdx++) { |
| 246 | sprintf(key, keyFmt, dwiIdx+nexIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), keyFmt, dwiIdx+nexIdx); |
| 247 | val = nrrdKeyValueGet(ndwi, key); |
| 248 | if (val) { |
| 249 | val = (char *)airFree(val); |
| 250 | biffAddf(TENtenBiffKey, "%s: shouldn't have key \"%s\" with " |
| 251 | "NEX %d for DWI %d", me, key, nexNum, dwiIdx); |
| 252 | airMopError(mop); return 1; |
| 253 | } |
| 254 | for (valIdx=0; valIdx<valNum; valIdx++) { |
| 255 | info[valIdx + valNum*nexIdx] = info[valIdx]; |
| 256 | } |
| 257 | } |
| 258 | dwiIdx += nexNum-1; |
| 259 | } |
| 260 | } |
| 261 | info += valNum*nexNum; |
| 262 | } |
| 263 | |
| 264 | /* perhaps too paranoid: see if there are extra keys, |
| 265 | which probably implies confusion/mismatch between number of |
| 266 | gradients and number of values */ |
| 267 | sprintf(key, keyFmt, dwiIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), keyFmt, dwiIdx); |
| 268 | val = nrrdKeyValueGet(ndwi, key); |
| 269 | if (val) { |
| 270 | biffAddf(TENtenBiffKey, "%s: saw \"%s\" key, more than required %u keys, " |
| 271 | "likely mismatch between keys and actual gradients", |
| 272 | me, key, dwiIdx); |
| 273 | airMopError(mop); return 1; |
| 274 | } |
| 275 | |
| 276 | /* second pass: see which ones are skipped, even though gradient/B-matrix |
| 277 | information has been specified */ |
| 278 | for (dwiIdx=0; dwiIdx<dwiNum; dwiIdx++) { |
| 279 | sprintf(key, tenDWMRISkipKeyFmt, dwiIdx)__builtin___sprintf_chk (key, 0, __builtin_object_size (key, 2 > 1 ? 1 : 0), tenDWMRISkipKeyFmt, dwiIdx); |
| 280 | val = nrrdKeyValueGet(ndwi, key); |
| 281 | if (val) { |
| 282 | airToLower(val); |
| 283 | if (!strncmp("true", val + strspn(val, AIR_WHITESPACE" \t\n\r\v\f"), |
| 284 | strlen("true"))) { |
| 285 | skipIdx = airArrayLenIncr(skipArr, 1); |
| 286 | (*skipP)[skipIdx] = dwiIdx; |
| 287 | skipLut[dwiIdx] = AIR_TRUE1; |
| 288 | } |
| 289 | } |
| 290 | } |
| 291 | |
| 292 | /* normalize so that maximal norm is 1.0 */ |
| 293 | /* Thu Dec 20 03:25:20 CST 2012 this rescaling IS in fact what is |
| 294 | causing the small discrepency between ngrad before and after |
| 295 | saving to KVPs. The problem is not related to how the gradient |
| 296 | vector coefficients are recovered from the string-based |
| 297 | representation; that is likely bit-for-bit correct. The problem |
| 298 | is when everything is rescaled by 1.0/normMax: a "normalized" |
| 299 | vector will not have *exactly* length 1.0. So what can be done |
| 300 | to prevent pointlessly altering the lengths of vectors that were |
| 301 | close enough to unit-length? Is there some more 754-savvy |
| 302 | way of doing this normalization? */ |
| 303 | normMax = 0; |
| 304 | info = (double *)(ninfo->data); |
| 305 | for (dwiIdx=0; dwiIdx<dwiNum; dwiIdx++) { |
| 306 | if (!skipLut[dwiIdx]) { |
| 307 | if (3 == valNum) { |
| 308 | norm = ELL_3V_LEN(info)(sqrt((((info))[0]*((info))[0] + ((info))[1]*((info))[1] + (( info))[2]*((info))[2]))); |
| 309 | } else { |
| 310 | norm = sqrt(info[0]*info[0] + 2*info[1]*info[1] + 2*info[2]*info[2] |
| 311 | + info[3]*info[3] + 2*info[4]*info[4] |
| 312 | + info[5]*info[5]); |
| 313 | } |
| 314 | normMax = AIR_MAX(normMax, norm)((normMax) > (norm) ? (normMax) : (norm)); |
| 315 | } |
| 316 | info += valNum; |
| 317 | } |
| 318 | info = (double *)(ninfo->data); |
| 319 | for (dwiIdx=0; dwiIdx<dwiNum; dwiIdx++) { |
| 320 | if (!skipLut[dwiIdx]) { |
| 321 | if (3 == valNum) { |
| 322 | ELL_3V_SCALE(info, 1.0/normMax, info)((info)[0] = (1.0/normMax)*(info)[0], (info)[1] = (1.0/normMax )*(info)[1], (info)[2] = (1.0/normMax)*(info)[2]); |
| 323 | } else { |
| 324 | ELL_6V_SCALE(info, 1.0/normMax, info)((info)[0] = (1.0/normMax)*(info)[0], (info)[1] = (1.0/normMax )*(info)[1], (info)[2] = (1.0/normMax)*(info)[2], (info)[3] = (1.0/normMax)*(info)[3], (info)[4] = (1.0/normMax)*(info)[4] , (info)[5] = (1.0/normMax)*(info)[5]); |
| 325 | } |
| 326 | } |
| 327 | info += valNum; |
| 328 | } |
| 329 | |
| 330 | airMopOkay(mop); |
| 331 | return 0; |
| 332 | } |
| 333 | |
| 334 | /* |
| 335 | ******** tenBMatrixCalc |
| 336 | ** |
| 337 | ** given a list of gradient directions (arbitrary type), contructs the |
| 338 | ** B-matrix that records how each coefficient of the diffusion tensor |
| 339 | ** is weighted in the diffusion weighted images. Matrix will be a |
| 340 | ** 6-by-N 2D array of doubles. |
| 341 | ** |
| 342 | ** NOTE 1: The ordering of the elements in each row is (like the ordering |
| 343 | ** of the tensor elements in all of ten): |
| 344 | ** |
| 345 | ** Bxx Bxy Bxz Byy Byz Bzz |
| 346 | ** |
| 347 | ** NOTE 2: The off-diagonal elements are NOT pre-multiplied by two. |
| 348 | */ |
| 349 | int |
| 350 | tenBMatrixCalc(Nrrd *nbmat, const Nrrd *_ngrad) { |
| 351 | static const char me[]="tenBMatrixCalc"; |
| 352 | Nrrd *ngrad; |
| 353 | double *bmat, *G; |
| 354 | int DD, dd; |
| 355 | airArray *mop; |
| 356 | |
| 357 | if (!(nbmat && _ngrad && !tenGradientCheck(_ngrad, nrrdTypeDefault, 1))) { |
| 358 | biffAddf(TENtenBiffKey, "%s: got NULL pointer or invalid arg", me); |
| 359 | return 1; |
| 360 | } |
| 361 | mop = airMopNew(); |
| 362 | airMopAdd(mop, ngrad=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 363 | if (nrrdConvert(ngrad, _ngrad, nrrdTypeDouble) |
| 364 | || nrrdMaybeAlloc_va(nbmat, nrrdTypeDouble, 2, |
| 365 | AIR_CAST(size_t, 6)((size_t)(6)), ngrad->axis[1].size)) { |
| 366 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: trouble", me); |
| 367 | airMopError(mop); return 1; |
| 368 | } |
| 369 | |
| 370 | DD = ngrad->axis[1].size; |
| 371 | G = (double*)(ngrad->data); |
| 372 | bmat = (double*)(nbmat->data); |
| 373 | for (dd=0; dd<DD; dd++) { |
| 374 | ELL_6V_SET(bmat,((bmat)[0]=(G[0]*G[0]), (bmat)[1]=(G[0]*G[1]), (bmat)[2]=(G[0 ]*G[2]), (bmat)[3]=(G[1]*G[1]), (bmat)[4]=(G[1]*G[2]), (bmat) [5]=(G[2]*G[2])) |
| 375 | G[0]*G[0], G[0]*G[1], G[0]*G[2],((bmat)[0]=(G[0]*G[0]), (bmat)[1]=(G[0]*G[1]), (bmat)[2]=(G[0 ]*G[2]), (bmat)[3]=(G[1]*G[1]), (bmat)[4]=(G[1]*G[2]), (bmat) [5]=(G[2]*G[2])) |
| 376 | G[1]*G[1], G[1]*G[2],((bmat)[0]=(G[0]*G[0]), (bmat)[1]=(G[0]*G[1]), (bmat)[2]=(G[0 ]*G[2]), (bmat)[3]=(G[1]*G[1]), (bmat)[4]=(G[1]*G[2]), (bmat) [5]=(G[2]*G[2])) |
| 377 | G[2]*G[2])((bmat)[0]=(G[0]*G[0]), (bmat)[1]=(G[0]*G[1]), (bmat)[2]=(G[0 ]*G[2]), (bmat)[3]=(G[1]*G[1]), (bmat)[4]=(G[1]*G[2]), (bmat) [5]=(G[2]*G[2])); |
| 378 | G += 3; |
| 379 | bmat += 6; |
| 380 | } |
| 381 | nbmat->axis[0].kind = nrrdKind3DSymMatrix; |
| 382 | |
| 383 | airMopOkay(mop); |
| 384 | return 0; |
| 385 | } |
| 386 | |
| 387 | /* |
| 388 | ******** tenEMatrixCalc |
| 389 | ** |
| 390 | */ |
| 391 | int |
| 392 | tenEMatrixCalc(Nrrd *nemat, const Nrrd *_nbmat, int knownB0) { |
| 393 | static const char me[]="tenEMatrixCalc"; |
| 394 | Nrrd *nbmat, *ntmp; |
| 395 | airArray *mop; |
| 396 | ptrdiff_t padmin[2], padmax[2]; |
| 397 | unsigned int ri; |
| 398 | double *bmat; |
| 399 | |
| 400 | if (!(nemat && _nbmat)) { |
| 401 | biffAddf(TENtenBiffKey, "%s: got NULL pointer", me); |
| 402 | return 1; |
| 403 | } |
| 404 | if (tenBMatrixCheck(_nbmat, nrrdTypeDefault, 6)) { |
| 405 | biffAddf(TENtenBiffKey, "%s: problem with B matrix", me); |
| 406 | return 1; |
| 407 | } |
| 408 | mop = airMopNew(); |
| 409 | airMopAdd(mop, nbmat=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 410 | if (knownB0) { |
| 411 | if (nrrdConvert(nbmat, _nbmat, nrrdTypeDouble)) { |
| 412 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't convert given bmat to doubles", me); |
| 413 | airMopError(mop); return 1; |
| 414 | } |
| 415 | } else { |
| 416 | airMopAdd(mop, ntmp=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 417 | if (nrrdConvert(ntmp, _nbmat, nrrdTypeDouble)) { |
| 418 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't convert given bmat to doubles", me); |
| 419 | airMopError(mop); return 1; |
| 420 | } |
| 421 | ELL_2V_SET(padmin, 0, 0)((padmin)[0]=(0), (padmin)[1]=(0)); |
| 422 | ELL_2V_SET(padmax, 6, _nbmat->axis[1].size-1)((padmax)[0]=(6), (padmax)[1]=(_nbmat->axis[1].size-1)); |
| 423 | if (nrrdPad_nva(nbmat, ntmp, padmin, padmax, nrrdBoundaryPad, -1)) { |
| 424 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't pad given bmat", me); |
| 425 | airMopError(mop); return 1; |
| 426 | } |
| 427 | } |
| 428 | bmat = (double*)(nbmat->data); |
| 429 | /* HERE is where the off-diagonal elements get multiplied by 2 */ |
| 430 | for (ri=0; ri<nbmat->axis[1].size; ri++) { |
| 431 | bmat[1] *= 2; |
| 432 | bmat[2] *= 2; |
| 433 | bmat[4] *= 2; |
| 434 | bmat += nbmat->axis[0].size; |
| 435 | } |
| 436 | if (ell_Nm_pseudo_inv(nemat, nbmat)) { |
| 437 | biffMovef(TENtenBiffKey, ELLell_biff_key, "%s: trouble pseudo-inverting B-matrix", me); |
| 438 | airMopError(mop); return 1; |
| 439 | } |
| 440 | airMopOkay(mop); |
| 441 | return 0; |
| 442 | } |
| 443 | |
| 444 | /* |
| 445 | ******** tenEstimateLinearSingle_d |
| 446 | ** |
| 447 | ** estimate one single tensor |
| 448 | ** |
| 449 | ** !! requires being passed a pre-allocated double array "vbuf" which is |
| 450 | ** !! used for intermediate calculations (details below) |
| 451 | ** |
| 452 | ** DD is always the length of the dwi[] array |
| 453 | ** |
| 454 | ** -------------- IF knownB0 ------------------------- |
| 455 | ** input: |
| 456 | ** dwi[0] is the B0 image, dwi[1]..dwi[DD-1] are the (DD-1) DWI values, |
| 457 | ** emat is the (DD-1)-by-6 estimation matrix, which is the pseudo-inverse |
| 458 | ** of the B-matrix (after the off-diagonals have been multiplied by 2). |
| 459 | ** vbuf[] is allocated for (at least) DD-1 doubles (DD is fine) |
| 460 | ** |
| 461 | ** output: |
| 462 | ** ten[0]..ten[6] will be the confidence value followed by the tensor |
| 463 | ** if B0P, then *B0P is set to the B0 value used in calcs: max(b0,1) |
| 464 | ** -------------- IF !knownB0 ------------------------- |
| 465 | ** input: |
| 466 | ** dwi[0]..dwi[DD-1] are the DD DWI values, emat is the DD-by-7 estimation |
| 467 | ** matrix. The 7th column is for estimating the B0 image. |
| 468 | ** vbuf[] is allocated for DD doubles |
| 469 | ** |
| 470 | ** output: |
| 471 | ** ten[0]..ten[6] will be the confidence value followed by the tensor |
| 472 | ** if B0P, then *B0P is set to estimated B0 value. |
| 473 | ** ---------------------------------------------------- |
| 474 | */ |
| 475 | void |
| 476 | tenEstimateLinearSingle_d(double *ten, double *B0P, /* output */ |
| 477 | const double *dwi, const double *emat, /* input .. */ |
| 478 | double *vbuf, unsigned int DD, int knownB0, |
| 479 | double thresh, double soft, double b) { |
| 480 | double logB0, tmp, mean; |
| 481 | unsigned int ii, jj; |
| 482 | /* static const char me[]="tenEstimateLinearSingle_d"; */ |
| 483 | |
| 484 | if (knownB0) { |
| 485 | if (B0P) { |
| 486 | /* we save this as a courtesy */ |
| 487 | *B0P = AIR_MAX(dwi[0], 1)((dwi[0]) > (1) ? (dwi[0]) : (1)); |
| 488 | } |
| 489 | logB0 = log(AIR_MAX(dwi[0], 1)((dwi[0]) > (1) ? (dwi[0]) : (1))); |
| 490 | mean = 0; |
| 491 | for (ii=1; ii<DD; ii++) { |
| 492 | tmp = AIR_MAX(dwi[ii], 1)((dwi[ii]) > (1) ? (dwi[ii]) : (1)); |
| 493 | mean += tmp; |
| 494 | vbuf[ii-1] = (logB0 - log(tmp))/b; |
| 495 | /* if (tenVerbose) { |
| 496 | fprintf(stderr, "vbuf[%d] = %f\n", ii-1, vbuf[ii-1]); |
| 497 | } */ |
| 498 | } |
| 499 | mean /= DD-1; |
| 500 | if (soft) { |
| 501 | ten[0] = AIR_AFFINE(-1, airErf((mean - thresh)/(soft + 0.000001)), 1,( ((double)(1)-(0))*((double)(airErf((mean - thresh)/(soft + 0.000001 )))-(-1)) / ((double)(1)-(-1)) + (0)) |
| 502 | 0, 1)( ((double)(1)-(0))*((double)(airErf((mean - thresh)/(soft + 0.000001 )))-(-1)) / ((double)(1)-(-1)) + (0)); |
| 503 | } else { |
| 504 | ten[0] = mean > thresh; |
| 505 | } |
| 506 | for (jj=0; jj<6; jj++) { |
| 507 | tmp = 0; |
| 508 | for (ii=0; ii<DD-1; ii++) { |
| 509 | tmp += emat[ii + (DD-1)*jj]*vbuf[ii]; |
| 510 | } |
| 511 | ten[jj+1] = tmp; |
| 512 | } |
| 513 | } else { |
| 514 | /* !knownB0 */ |
| 515 | mean = 0; |
| 516 | for (ii=0; ii<DD; ii++) { |
| 517 | tmp = AIR_MAX(dwi[ii], 1)((dwi[ii]) > (1) ? (dwi[ii]) : (1)); |
| 518 | mean += tmp; |
| 519 | vbuf[ii] = -log(tmp)/b; |
| 520 | } |
| 521 | mean /= DD; |
| 522 | if (soft) { |
| 523 | ten[0] = AIR_AFFINE(-1, airErf((mean - thresh)/(soft + 0.000001)), 1,( ((double)(1)-(0))*((double)(airErf((mean - thresh)/(soft + 0.000001 )))-(-1)) / ((double)(1)-(-1)) + (0)) |
| 524 | 0, 1)( ((double)(1)-(0))*((double)(airErf((mean - thresh)/(soft + 0.000001 )))-(-1)) / ((double)(1)-(-1)) + (0)); |
| 525 | } else { |
| 526 | ten[0] = mean > thresh; |
| 527 | } |
| 528 | for (jj=0; jj<7; jj++) { |
| 529 | tmp = 0; |
| 530 | for (ii=0; ii<DD; ii++) { |
| 531 | tmp += emat[ii + DD*jj]*vbuf[ii]; |
| 532 | } |
| 533 | if (jj < 6) { |
| 534 | ten[jj+1] = tmp; |
| 535 | } else { |
| 536 | /* we're on seventh row, for finding B0 */ |
| 537 | if (B0P) { |
| 538 | *B0P = exp(b*tmp); |
| 539 | } |
| 540 | } |
| 541 | } |
| 542 | } |
| 543 | return; |
| 544 | } |
| 545 | |
| 546 | void |
| 547 | tenEstimateLinearSingle_f(float *_ten, float *_B0P, /* output */ |
| 548 | const float *_dwi, const double *emat, /* input .. */ |
| 549 | double *vbuf, unsigned int DD, int knownB0, |
| 550 | float thresh, float soft, float b) { |
| 551 | static const char me[]="tenEstimateLinearSingle_f"; |
| 552 | #define DWI_NUM_MAX256 256 |
| 553 | double dwi[DWI_NUM_MAX256], ten[7], B0; |
| 554 | unsigned int dwiIdx; |
| 555 | |
| 556 | /* HEY: this is somewhat inelegant .. */ |
| 557 | if (DD > DWI_NUM_MAX256) { |
| 558 | fprintf(stderr__stderrp, "%s: PANIC: sorry, DD=%u > compile-time DWI_NUM_MAX=%u\n", |
| 559 | me, DD, DWI_NUM_MAX256); |
| 560 | exit(1); |
| 561 | } |
| 562 | for (dwiIdx=0; dwiIdx<DD; dwiIdx++) { |
| 563 | dwi[dwiIdx] = _dwi[dwiIdx]; |
| 564 | } |
| 565 | tenEstimateLinearSingle_d(ten, _B0P ? &B0 : NULL((void*)0), |
| 566 | dwi, emat, |
| 567 | vbuf, DD, knownB0, |
| 568 | thresh, soft, b); |
| 569 | TEN_T_COPY_TT(_ten, float, ten)( (_ten)[0] = ((float)((ten)[0])), (_ten)[1] = ((float)((ten) [1])), (_ten)[2] = ((float)((ten)[2])), (_ten)[3] = ((float)( (ten)[3])), (_ten)[4] = ((float)((ten)[4])), (_ten)[5] = ((float )((ten)[5])), (_ten)[6] = ((float)((ten)[6])) ); |
| 570 | if (_B0P) { |
| 571 | *_B0P = AIR_CAST(float, B0)((float)(B0)); |
| 572 | } |
| 573 | return; |
| 574 | } |
| 575 | |
| 576 | /* |
| 577 | ******** tenEstimateLinear3D |
| 578 | ** |
| 579 | ** takes an array of DWIs (starting with the B=0 image), joins them up, |
| 580 | ** and passes it all off to tenEstimateLinear4D |
| 581 | ** |
| 582 | ** Note: this will copy per-axis peripheral information from _ndwi[0] |
| 583 | */ |
| 584 | int |
| 585 | tenEstimateLinear3D(Nrrd *nten, Nrrd **nterrP, Nrrd **nB0P, |
| 586 | const Nrrd *const *_ndwi, unsigned int dwiLen, |
| 587 | const Nrrd *_nbmat, int knownB0, |
| 588 | double thresh, double soft, double b) { |
| 589 | static const char me[]="tenEstimateLinear3D"; |
| 590 | Nrrd *ndwi; |
| 591 | airArray *mop; |
| 592 | int amap[4] = {-1, 0, 1, 2}; |
| 593 | |
| 594 | if (!(_ndwi)) { |
| 595 | biffAddf(TENtenBiffKey, "%s: got NULL pointer", me); |
| 596 | return 1; |
| 597 | } |
| 598 | mop = airMopNew(); |
| 599 | ndwi = nrrdNew(); |
| 600 | airMopAdd(mop, ndwi, (airMopper)nrrdNuke, airMopAlways); |
| 601 | if (nrrdJoin(ndwi, (const Nrrd*const*)_ndwi, dwiLen, 0, AIR_TRUE1)) { |
| 602 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: trouble joining inputs", me); |
| 603 | airMopError(mop); return 1; |
| 604 | } |
| 605 | |
| 606 | nrrdAxisInfoCopy(ndwi, _ndwi[0], amap, NRRD_AXIS_INFO_NONE0); |
| 607 | if (tenEstimateLinear4D(nten, nterrP, nB0P, |
| 608 | ndwi, _nbmat, knownB0, thresh, soft, b)) { |
| 609 | biffAddf(TENtenBiffKey, "%s: trouble", me); |
| 610 | airMopError(mop); return 1; |
| 611 | } |
| 612 | |
| 613 | airMopOkay(mop); |
| 614 | return 0; |
| 615 | } |
| 616 | |
| 617 | /* |
| 618 | ******** tenEstimateLinear4D |
| 619 | ** |
| 620 | ** given a stack of DWI volumes (ndwi) and the imaging B-matrix used |
| 621 | ** for acquisiton (_nbmat), computes and stores diffusion tensors in |
| 622 | ** nten. |
| 623 | ** |
| 624 | ** The mean of the diffusion-weighted images is thresholded at "thresh" with |
| 625 | ** softness parameter "soft". |
| 626 | ** |
| 627 | ** This takes the B-matrix (weighting matrix), such as formed by tenBMatrix, |
| 628 | ** or from a more complete account of the gradients present in an imaging |
| 629 | ** sequence, and then does the pseudo inverse to get the estimation matrix |
| 630 | */ |
| 631 | int |
| 632 | tenEstimateLinear4D(Nrrd *nten, Nrrd **nterrP, Nrrd **nB0P, |
| 633 | const Nrrd *ndwi, const Nrrd *_nbmat, int knownB0, |
| 634 | double thresh, double soft, double b) { |
| 635 | static const char me[]="tenEstimateLinear4D"; |
| 636 | Nrrd *nemat, *nbmat, *ncrop, *nhist; |
| 637 | airArray *mop; |
| 638 | size_t cmin[4], cmax[4], sx, sy, sz, II, d, DD; |
| 639 | int E, amap[4]; |
| 640 | float *ten, *dwi1, *dwi2, *terr, |
| 641 | _B0, *B0, (*lup)(const void *, size_t); |
| 642 | double *emat, *bmat, *vbuf; |
| 643 | NrrdRange *range; |
| 644 | float te, d1, d2; |
| 645 | char stmp[2][AIR_STRLEN_SMALL(128+1)]; |
| 646 | |
| 647 | if (!(nten && ndwi && _nbmat)) { |
| 648 | /* nerrP and _NB0P can be NULL */ |
| 649 | biffAddf(TENtenBiffKey, "%s: got NULL pointer", me); |
| 650 | return 1; |
| 651 | } |
| 652 | if (!( 4 == ndwi->dim && 7 <= ndwi->axis[0].size )) { |
| 653 | biffAddf(TENtenBiffKey, "%s: dwi should be 4-D array with axis 0 size >= 7", me); |
| 654 | return 1; |
| 655 | } |
| 656 | if (tenBMatrixCheck(_nbmat, nrrdTypeDefault, 6)) { |
| 657 | biffAddf(TENtenBiffKey, "%s: problem with B matrix", me); |
| 658 | return 1; |
| 659 | } |
| 660 | if (knownB0) { |
| 661 | if (!( ndwi->axis[0].size == 1 + _nbmat->axis[1].size )) { |
| 662 | biffAddf(TENtenBiffKey, "%s: (knownB0 == true) # input images (%s) " |
| 663 | "!= 1 + # B matrix rows (1+%s)", me, |
| 664 | airSprintSize_t(stmp[0], ndwi->axis[0].size), |
| 665 | airSprintSize_t(stmp[1], _nbmat->axis[1].size)); |
| 666 | return 1; |
| 667 | } |
| 668 | } else { |
| 669 | if (!( ndwi->axis[0].size == _nbmat->axis[1].size )) { |
| 670 | biffAddf(TENtenBiffKey, "%s: (knownB0 == false) # dwi (%s) " |
| 671 | "!= # B matrix rows (%s)", me, |
| 672 | airSprintSize_t(stmp[0], ndwi->axis[0].size), |
| 673 | airSprintSize_t(stmp[1], _nbmat->axis[1].size)); |
| 674 | return 1; |
| 675 | } |
| 676 | } |
| 677 | |
| 678 | DD = ndwi->axis[0].size; |
| 679 | sx = ndwi->axis[1].size; |
| 680 | sy = ndwi->axis[2].size; |
| 681 | sz = ndwi->axis[3].size; |
| 682 | mop = airMopNew(); |
| 683 | airMopAdd(mop, nbmat=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 684 | if (nrrdConvert(nbmat, _nbmat, nrrdTypeDouble)) { |
| 685 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: trouble converting to doubles", me); |
| 686 | airMopError(mop); return 1; |
| 687 | } |
| 688 | airMopAdd(mop, nemat=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 689 | if (tenEMatrixCalc(nemat, nbmat, knownB0)) { |
| 690 | biffAddf(TENtenBiffKey, "%s: trouble computing estimation matrix", me); |
| 691 | airMopError(mop); return 1; |
| 692 | } |
| 693 | vbuf = AIR_CALLOC(knownB0 ? DD-1 : DD, double)(double*)(calloc((knownB0 ? DD-1 : DD), sizeof(double))); |
| 694 | dwi1 = AIR_CALLOC(DD, float)(float*)(calloc((DD), sizeof(float))); |
| 695 | dwi2 = AIR_CALLOC(knownB0 ? DD : DD+1, float)(float*)(calloc((knownB0 ? DD : DD+1), sizeof(float))); |
| 696 | airMopAdd(mop, vbuf, airFree, airMopAlways); |
| 697 | airMopAdd(mop, dwi1, airFree, airMopAlways); |
| 698 | airMopAdd(mop, dwi2, airFree, airMopAlways); |
| 699 | if (!(vbuf && dwi1 && dwi2)) { |
| 700 | biffAddf(TENtenBiffKey, "%s: couldn't allocate temp buffers", me); |
| 701 | airMopError(mop); return 1; |
| 702 | } |
| 703 | if (!AIR_EXISTS(thresh)(((int)(!((thresh) - (thresh)))))) { |
| 704 | airMopAdd(mop, ncrop=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 705 | airMopAdd(mop, nhist=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 706 | ELL_4V_SET(cmin, knownB0 ? 1 : 0, 0, 0, 0)((cmin)[0] = (knownB0 ? 1 : 0), (cmin)[1] = (0), (cmin)[2] = ( 0), (cmin)[3] = (0)); |
| 707 | ELL_4V_SET(cmax, DD-1, sx-1, sy-1, sz-1)((cmax)[0] = (DD-1), (cmax)[1] = (sx-1), (cmax)[2] = (sy-1), ( cmax)[3] = (sz-1)); |
| 708 | E = 0; |
| 709 | if (!E) E |= nrrdCrop(ncrop, ndwi, cmin, cmax); |
| 710 | if (!E) range = nrrdRangeNewSet(ncrop, nrrdBlind8BitRangeState); |
| 711 | if (!E) airMopAdd(mop, range, (airMopper)nrrdRangeNix, airMopAlways); |
| 712 | if (!E) E |= nrrdHisto(nhist, ncrop, range, NULL((void*)0), |
| 713 | (int)AIR_MIN(1024, range->max - range->min + 1)((1024) < (range->max - range->min + 1) ? (1024) : ( range->max - range->min + 1)), |
| 714 | nrrdTypeFloat); |
| 715 | if (E) { |
| 716 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, |
| 717 | "%s: trouble histograming to find DW threshold", me); |
| 718 | airMopError(mop); return 1; |
| 719 | } |
| 720 | if (_tenFindValley(&thresh, nhist, 0.75, AIR_FALSE0)) { |
| 721 | biffAddf(TENtenBiffKey, "%s: problem finding DW histogram valley", me); |
| 722 | airMopError(mop); return 1; |
| 723 | } |
| 724 | fprintf(stderr__stderrp, "%s: using %g for DW confidence threshold\n", me, thresh); |
| 725 | } |
| 726 | if (nrrdMaybeAlloc_va(nten, nrrdTypeFloat, 4, |
| 727 | AIR_CAST(size_t, 7)((size_t)(7)), sx, sy, sz)) { |
| 728 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't allocate output", me); |
| 729 | airMopError(mop); return 1; |
| 730 | } |
| 731 | if (nterrP) { |
| 732 | if (!(*nterrP)) { |
| 733 | *nterrP = nrrdNew(); |
| 734 | } |
| 735 | if (nrrdMaybeAlloc_va(*nterrP, nrrdTypeFloat, 3, |
| 736 | sx, sy, sz)) { |
| 737 | biffAddf(NRRDnrrdBiffKey, "%s: couldn't allocate error output", me); |
| 738 | airMopError(mop); return 1; |
| 739 | } |
| 740 | airMopAdd(mop, nterrP, (airMopper)airSetNull, airMopOnError); |
| 741 | airMopAdd(mop, *nterrP, (airMopper)nrrdNuke, airMopOnError); |
| 742 | terr = (float*)((*nterrP)->data); |
| 743 | } else { |
| 744 | terr = NULL((void*)0); |
| 745 | } |
| 746 | if (nB0P) { |
| 747 | if (!(*nB0P)) { |
| 748 | *nB0P = nrrdNew(); |
| 749 | } |
| 750 | if (nrrdMaybeAlloc_va(*nB0P, nrrdTypeFloat, 3, |
| 751 | sx, sy, sz)) { |
| 752 | biffAddf(NRRDnrrdBiffKey, "%s: couldn't allocate error output", me); |
| 753 | airMopError(mop); return 1; |
| 754 | } |
| 755 | airMopAdd(mop, nB0P, (airMopper)airSetNull, airMopOnError); |
| 756 | airMopAdd(mop, *nB0P, (airMopper)nrrdNuke, airMopOnError); |
| 757 | B0 = (float*)((*nB0P)->data); |
| 758 | } else { |
| 759 | B0 = NULL((void*)0); |
| 760 | } |
| 761 | bmat = (double*)(nbmat->data); |
| 762 | emat = (double*)(nemat->data); |
| 763 | ten = (float*)(nten->data); |
| 764 | lup = nrrdFLookup[ndwi->type]; |
| 765 | for (II=0; II<sx*sy*sz; II++) { |
| 766 | /* tenVerbose = (II == 42 + 190*(96 + 196*0)); */ |
| 767 | for (d=0; d<DD; d++) { |
| 768 | dwi1[d] = lup(ndwi->data, d + DD*II); |
| 769 | /* if (tenVerbose) { |
| 770 | fprintf(stderr, "%s: input dwi1[%d] = %g\n", me, d, dwi1[d]); |
| 771 | } */ |
| 772 | } |
| 773 | tenEstimateLinearSingle_f(ten, &_B0, dwi1, emat, |
| 774 | vbuf, DD, knownB0, |
| 775 | AIR_CAST(float, thresh)((float)(thresh)), |
| 776 | AIR_CAST(float, soft)((float)(soft)), |
| 777 | AIR_CAST(float, b)((float)(b))); |
| 778 | if (nB0P) { |
| 779 | *B0 = _B0; |
| 780 | } |
| 781 | /* if (tenVerbose) { |
| 782 | fprintf(stderr, "%s: output ten = (%g) %g,%g,%g %g,%g %g\n", me, |
| 783 | ten[0], ten[1], ten[2], ten[3], ten[4], ten[5], ten[6]); |
| 784 | } */ |
| 785 | if (nterrP) { |
| 786 | te = 0; |
| 787 | if (knownB0) { |
| 788 | tenSimulateSingle_f(dwi2, _B0, ten, bmat, DD, AIR_CAST(float, b)((float)(b))); |
| 789 | for (d=1; d<DD; d++) { |
| 790 | d1 = AIR_MAX(dwi1[d], 1)((dwi1[d]) > (1) ? (dwi1[d]) : (1)); |
| 791 | d2 = AIR_MAX(dwi2[d], 1)((dwi2[d]) > (1) ? (dwi2[d]) : (1)); |
| 792 | te += (d1 - d2)*(d1 - d2); |
| 793 | } |
| 794 | te /= (DD-1); |
| 795 | } else { |
| 796 | tenSimulateSingle_f(dwi2, _B0, ten, bmat, DD+1, AIR_CAST(float, b)((float)(b))); |
| 797 | for (d=0; d<DD; d++) { |
| 798 | d1 = AIR_MAX(dwi1[d], 1)((dwi1[d]) > (1) ? (dwi1[d]) : (1)); |
| 799 | /* tenSimulateSingle_f always puts the B0 in the beginning of |
| 800 | the dwi vector, but in this case we didn't have it in |
| 801 | the input dwi vector */ |
| 802 | d2 = AIR_MAX(dwi2[d+1], 1)((dwi2[d+1]) > (1) ? (dwi2[d+1]) : (1)); |
| 803 | te += (d1 - d2)*(d1 - d2); |
| 804 | } |
| 805 | te /= DD; |
| 806 | } |
| 807 | *terr = AIR_CAST(float, sqrt(te))((float)(sqrt(te))); |
| 808 | terr += 1; |
| 809 | } |
| 810 | ten += 7; |
| 811 | if (B0) { |
| 812 | B0 += 1; |
| 813 | } |
| 814 | } |
| 815 | /* not our job: tenEigenvalueClamp(nten, nten, 0, AIR_NAN); */ |
| 816 | |
| 817 | ELL_4V_SET(amap, -1, 1, 2, 3)((amap)[0] = (-1), (amap)[1] = (1), (amap)[2] = (2), (amap)[3 ] = (3)); |
| 818 | nrrdAxisInfoCopy(nten, ndwi, amap, NRRD_AXIS_INFO_NONE0); |
| 819 | nten->axis[0].kind = nrrdKind3DMaskedSymMatrix; |
| 820 | if (nrrdBasicInfoCopy(nten, ndwi, |
| 821 | NRRD_BASIC_INFO_ALL((1<<1)|(1<<2)|(1<<3)|(1<<4)|(1<< 5)|(1<<6)|(1<<7)|(1<<8)|(1<<9)|(1<< 10) |(1<<11)|(1<<12)|(1<<13)|(1<<14)| (1<<15)) ^ NRRD_BASIC_INFO_SPACE((1<< 7) | (1<< 8) | (1<< 9) | (1<<10 ) | (1<<11)))) { |
| 822 | biffAddf(NRRDnrrdBiffKey, "%s:", me); |
| 823 | return 1; |
| 824 | } |
| 825 | |
| 826 | airMopOkay(mop); |
| 827 | return 0; |
| 828 | } |
| 829 | |
| 830 | /* |
| 831 | ******** tenSimulateSingle_f |
| 832 | ** |
| 833 | ** given a tensor, simulate the set of diffusion weighted measurements |
| 834 | ** represented by the given B matrix |
| 835 | ** |
| 836 | ** NOTE: the mindset of this function is very much "knownB0==true": |
| 837 | ** B0 is required as an argument (and its always copied to dwi[0]), |
| 838 | ** and the given bmat is assumed to have DD-1 rows (similar to how |
| 839 | ** tenEstimateLinearSingle_f() is set up), and dwi[1] through dwi[DD-1] |
| 840 | ** are set to the calculated DWIs. |
| 841 | ** |
| 842 | ** So: dwi must be allocated for DD values total |
| 843 | */ |
| 844 | void |
| 845 | tenSimulateSingle_f(float *dwi, |
| 846 | float B0, const float *ten, const double *bmat, |
| 847 | unsigned int DD, float b) { |
| 848 | double vv; |
| 849 | /* this is how we multiply the off-diagonal entries by 2 */ |
| 850 | double matwght[6] = {1, 2, 2, 1, 2, 1}; |
| 851 | unsigned int ii, jj; |
| 852 | |
| 853 | dwi[0] = B0; |
| 854 | /* if (tenVerbose) { |
| 855 | fprintf(stderr, "ten = %g,%g,%g %g,%g %g\n", |
| 856 | ten[1], ten[2], ten[3], ten[4], ten[5], ten[6]); |
| 857 | } */ |
| 858 | for (ii=0; ii<DD-1; ii++) { |
| 859 | vv = 0; |
| 860 | for (jj=0; jj<6; jj++) { |
| 861 | vv += matwght[jj]*bmat[jj + 6*ii]*ten[jj+1]; |
| 862 | } |
| 863 | dwi[ii+1] = AIR_CAST(float, AIR_MAX(B0, 1)*exp(-b*vv))((float)(((B0) > (1) ? (B0) : (1))*exp(-b*vv))); |
| 864 | /* if (tenVerbose) { |
| 865 | fprintf(stderr, "v[%d] = %g --> dwi = %g\n", ii, vv, dwi[ii+1]); |
| 866 | } */ |
| 867 | } |
| 868 | |
| 869 | return; |
| 870 | } |
| 871 | |
| 872 | int |
| 873 | tenSimulate(Nrrd *ndwi, const Nrrd *nT2, const Nrrd *nten, |
| 874 | const Nrrd *_nbmat, double b) { |
| 875 | static const char me[]="tenSimulate"; |
| 876 | size_t II; |
| 877 | Nrrd *nbmat; |
| 878 | size_t DD, sx, sy, sz; |
| 879 | airArray *mop; |
| 880 | double *bmat; |
| 881 | float *dwi, *ten, (*lup)(const void *, size_t I); |
| 882 | char stmp[6][AIR_STRLEN_SMALL(128+1)]; |
| 883 | |
| 884 | if (!ndwi || !nT2 || !nten || !_nbmat |
| 885 | || tenTensorCheck(nten, nrrdTypeFloat, AIR_TRUE1, AIR_TRUE1) |
| 886 | || tenBMatrixCheck(_nbmat, nrrdTypeDefault, 6)) { |
| 887 | biffAddf(TENtenBiffKey, "%s: got NULL pointer or invalid args", me); |
| 888 | return 1; |
| 889 | } |
| 890 | mop = airMopNew(); |
| 891 | airMopAdd(mop, nbmat=nrrdNew(), (airMopper)nrrdNuke, airMopAlways); |
| 892 | if (nrrdConvert(nbmat, _nbmat, nrrdTypeDouble)) { |
| 893 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't convert B matrix", me); |
| 894 | return 1; |
| 895 | } |
| 896 | |
| 897 | DD = nbmat->axis[1].size+1; |
| 898 | sx = nT2->axis[0].size; |
| 899 | sy = nT2->axis[1].size; |
| 900 | sz = nT2->axis[2].size; |
| 901 | if (!(3 == nT2->dim |
| 902 | && sx == nten->axis[1].size |
| 903 | && sy == nten->axis[2].size |
| 904 | && sz == nten->axis[3].size)) { |
| 905 | biffAddf(TENtenBiffKey, "%s: dimensions of %u-D T2 volume (%s,%s,%s) " |
| 906 | "don't match tensor volume (%s,%s,%s)", me, nT2->dim, |
| 907 | airSprintSize_t(stmp[0], sx), |
| 908 | airSprintSize_t(stmp[1], sy), |
| 909 | airSprintSize_t(stmp[2], sz), |
| 910 | airSprintSize_t(stmp[3], nten->axis[1].size), |
| 911 | airSprintSize_t(stmp[4], nten->axis[2].size), |
| 912 | airSprintSize_t(stmp[5], nten->axis[3].size)); |
| 913 | return 1; |
| 914 | } |
| 915 | if (nrrdMaybeAlloc_va(ndwi, nrrdTypeFloat, 4, |
| 916 | AIR_CAST(size_t, DD)((size_t)(DD)), sx, sy, sz)) { |
| 917 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't allocate output", me); |
| 918 | return 1; |
| 919 | } |
| 920 | dwi = (float*)(ndwi->data); |
| 921 | ten = (float*)(nten->data); |
| 922 | bmat = (double*)(nbmat->data); |
| 923 | lup = nrrdFLookup[nT2->type]; |
| 924 | for (II=0; II<(size_t)(sx*sy*sz); II++) { |
| 925 | /* tenVerbose = (II == 42 + 190*(96 + 196*0)); */ |
| 926 | tenSimulateSingle_f(dwi, lup(nT2->data, II), ten, bmat, DD, |
| 927 | AIR_CAST(float, b)((float)(b))); |
| 928 | dwi += DD; |
| 929 | ten += 7; |
| 930 | } |
| 931 | |
| 932 | airMopOkay(mop); |
| 933 | return 0; |
| 934 | } |
| 935 | |
| 936 | |
| 937 | |
| 938 | |
| 939 | |
| 940 | |
| 941 | |
| 942 | |
| 943 | |
| 944 | |
| 945 | |
| 946 | |
| 947 | |
| 948 | |
| 949 | |
| 950 | |
| 951 | |
| 952 | |
| 953 | /* old stuff, prior to tenEstimationMatrix .. */ |
| 954 | |
| 955 | |
| 956 | /* |
| 957 | ******** tenCalcOneTensor1 |
| 958 | ** |
| 959 | ** make one diffusion tensor from the measurements at one voxel, based |
| 960 | ** on the gradient directions used by Andy Alexander |
| 961 | */ |
| 962 | void |
| 963 | tenCalcOneTensor1(float tens[7], float chan[7], |
| 964 | float thresh, float slope, float b) { |
| 965 | double c[7], sum, d1, d2, d3, d4, d5, d6; |
| 966 | |
| 967 | c[0] = AIR_MAX(chan[0], 1)((chan[0]) > (1) ? (chan[0]) : (1)); |
| 968 | c[1] = AIR_MAX(chan[1], 1)((chan[1]) > (1) ? (chan[1]) : (1)); |
| 969 | c[2] = AIR_MAX(chan[2], 1)((chan[2]) > (1) ? (chan[2]) : (1)); |
| 970 | c[3] = AIR_MAX(chan[3], 1)((chan[3]) > (1) ? (chan[3]) : (1)); |
| 971 | c[4] = AIR_MAX(chan[4], 1)((chan[4]) > (1) ? (chan[4]) : (1)); |
| 972 | c[5] = AIR_MAX(chan[5], 1)((chan[5]) > (1) ? (chan[5]) : (1)); |
| 973 | c[6] = AIR_MAX(chan[6], 1)((chan[6]) > (1) ? (chan[6]) : (1)); |
| 974 | sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6]; |
| 975 | tens[0] = AIR_CAST(float, (1 + airErf(slope*(sum - thresh)))/2.0)((float)((1 + airErf(slope*(sum - thresh)))/2.0)); |
| 976 | d1 = (log(c[0]) - log(c[1]))/b; |
| 977 | d2 = (log(c[0]) - log(c[2]))/b; |
| 978 | d3 = (log(c[0]) - log(c[3]))/b; |
| 979 | d4 = (log(c[0]) - log(c[4]))/b; |
| 980 | d5 = (log(c[0]) - log(c[5]))/b; |
| 981 | d6 = (log(c[0]) - log(c[6]))/b; |
| 982 | tens[1] = AIR_CAST(float, d1 + d2 - d3 - d4 + d5 + d6)((float)(d1 + d2 - d3 - d4 + d5 + d6)); /* Dxx */ |
| 983 | tens[2] = AIR_CAST(float, d5 - d6)((float)(d5 - d6)); /* Dxy */ |
| 984 | tens[3] = AIR_CAST(float, d1 - d2)((float)(d1 - d2)); /* Dxz */ |
| 985 | tens[4] = AIR_CAST(float, -d1 - d2 + d3 + d4 + d5 + d6)((float)(-d1 - d2 + d3 + d4 + d5 + d6)); /* Dyy */ |
| 986 | tens[5] = AIR_CAST(float, d3 - d4)((float)(d3 - d4)); /* Dyz */ |
| 987 | tens[6] = AIR_CAST(float, d1 + d2 + d3 + d4 - d5 - d6)((float)(d1 + d2 + d3 + d4 - d5 - d6)); /* Dzz */ |
| 988 | return; |
| 989 | } |
| 990 | |
| 991 | /* |
| 992 | ******** tenCalcOneTensor2 |
| 993 | ** |
| 994 | ** using gradient directions used by EK |
| 995 | */ |
| 996 | void |
| 997 | tenCalcOneTensor2(float tens[7], float chan[7], |
| 998 | float thresh, float slope, float b) { |
| 999 | double c[7], sum, d1, d2, d3, d4, d5, d6; |
| 1000 | |
| 1001 | c[0] = AIR_MAX(chan[0], 1)((chan[0]) > (1) ? (chan[0]) : (1)); |
| 1002 | c[1] = AIR_MAX(chan[1], 1)((chan[1]) > (1) ? (chan[1]) : (1)); |
| 1003 | c[2] = AIR_MAX(chan[2], 1)((chan[2]) > (1) ? (chan[2]) : (1)); |
| 1004 | c[3] = AIR_MAX(chan[3], 1)((chan[3]) > (1) ? (chan[3]) : (1)); |
| 1005 | c[4] = AIR_MAX(chan[4], 1)((chan[4]) > (1) ? (chan[4]) : (1)); |
| 1006 | c[5] = AIR_MAX(chan[5], 1)((chan[5]) > (1) ? (chan[5]) : (1)); |
| 1007 | c[6] = AIR_MAX(chan[6], 1)((chan[6]) > (1) ? (chan[6]) : (1)); |
| 1008 | sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6]; |
| 1009 | tens[0] = AIR_CAST(float, (1 + airErf(slope*(sum - thresh)))/2.0)((float)((1 + airErf(slope*(sum - thresh)))/2.0)); |
| 1010 | d1 = (log(c[0]) - log(c[1]))/b; |
| 1011 | d2 = (log(c[0]) - log(c[2]))/b; |
| 1012 | d3 = (log(c[0]) - log(c[3]))/b; |
| 1013 | d4 = (log(c[0]) - log(c[4]))/b; |
| 1014 | d5 = (log(c[0]) - log(c[5]))/b; |
| 1015 | d6 = (log(c[0]) - log(c[6]))/b; |
| 1016 | tens[1] = AIR_CAST(float, d1)((float)(d1)); /* Dxx */ |
| 1017 | tens[2] = AIR_CAST(float, d6 - (d1 + d2)/2)((float)(d6 - (d1 + d2)/2)); /* Dxy */ |
| 1018 | tens[3] = AIR_CAST(float, d5 - (d1 + d3)/2)((float)(d5 - (d1 + d3)/2)); /* Dxz */ |
| 1019 | tens[4] = AIR_CAST(float, d2)((float)(d2)); /* Dyy */ |
| 1020 | tens[5] = AIR_CAST(float, d4 - (d2 + d3)/2)((float)(d4 - (d2 + d3)/2)); /* Dyz */ |
| 1021 | tens[6] = AIR_CAST(float, d3)((float)(d3)); /* Dzz */ |
| 1022 | return; |
| 1023 | } |
| 1024 | |
| 1025 | /* |
| 1026 | ******** tenCalcTensor |
| 1027 | ** |
| 1028 | ** Calculate a volume of tensors from measured data |
| 1029 | */ |
| 1030 | int |
| 1031 | tenCalcTensor(Nrrd *nout, Nrrd *nin, int version, |
| 1032 | float thresh, float slope, float b) { |
| 1033 | static const char me[] = "tenCalcTensor"; |
| 1034 | char cmt[128]; |
| 1035 | float *out, tens[7], chan[7]; |
| 1036 | size_t I, sx, sy, sz; |
| 1037 | void (*calcten)(float tens[7], float chan[7], |
| 1038 | float thresh, float slope, float b); |
| 1039 | |
| 1040 | if (!(nout && nin)) { |
| 1041 | biffAddf(TENtenBiffKey, "%s: got NULL pointer", me); |
| 1042 | return 1; |
| 1043 | } |
| 1044 | if (!( 1 == version || 2 == version )) { |
| 1045 | biffAddf(TENtenBiffKey, "%s: version should be 1 or 2, not %d", me, version); |
| 1046 | return 1; |
| 1047 | } |
| 1048 | switch (version) { |
| 1049 | case 1: |
| 1050 | calcten = tenCalcOneTensor1; |
| 1051 | break; |
| 1052 | case 2: |
| 1053 | calcten = tenCalcOneTensor2; |
| 1054 | break; |
| 1055 | default: |
| 1056 | biffAddf(TENtenBiffKey, "%s: PANIC, version = %d not handled", me, version); |
| 1057 | return 1; |
| 1058 | break; |
| 1059 | } |
| 1060 | if (tenTensorCheck(nin, nrrdTypeDefault, AIR_TRUE1, AIR_TRUE1)) { |
| 1061 | biffAddf(TENtenBiffKey, "%s: wasn't given valid tensor nrrd", me); |
| 1062 | return 1; |
| 1063 | } |
| 1064 | sx = nin->axis[1].size; |
| 1065 | sy = nin->axis[2].size; |
| 1066 | sz = nin->axis[3].size; |
| 1067 | if (nrrdMaybeAlloc_va(nout, nrrdTypeFloat, 4, |
| 1068 | AIR_CAST(size_t, 7)((size_t)(7)), sx, sy, sz)) { |
| 1069 | biffMovef(TENtenBiffKey, NRRDnrrdBiffKey, "%s: couldn't alloc output", me); |
| 1070 | return 1; |
| 1071 | } |
| 1072 | nout->axis[0].label = airStrdup("c,Dxx,Dxy,Dxz,Dyy,Dyz,Dzz"); |
| 1073 | nout->axis[1].label = airStrdup("x"); |
| 1074 | nout->axis[2].label = airStrdup("y"); |
| 1075 | nout->axis[3].label = airStrdup("z"); |
| 1076 | nout->axis[0].spacing = AIR_NAN(airFloatQNaN.f); |
| 1077 | if (AIR_EXISTS(nin->axis[1].spacing)(((int)(!((nin->axis[1].spacing) - (nin->axis[1].spacing ))))) && |
| 1078 | AIR_EXISTS(nin->axis[2].spacing)(((int)(!((nin->axis[2].spacing) - (nin->axis[2].spacing ))))) && |
| 1079 | AIR_EXISTS(nin->axis[3].spacing)(((int)(!((nin->axis[3].spacing) - (nin->axis[3].spacing )))))) { |
| 1080 | nout->axis[1].spacing = nin->axis[1].spacing; |
| 1081 | nout->axis[2].spacing = nin->axis[2].spacing; |
| 1082 | nout->axis[3].spacing = nin->axis[3].spacing; |
| 1083 | } |
| 1084 | else { |
| 1085 | nout->axis[1].spacing = 1.0; |
| 1086 | nout->axis[2].spacing = 1.0; |
| 1087 | nout->axis[3].spacing = 1.0; |
| 1088 | } |
| 1089 | sprintf(cmt, "%s: using thresh = %g, slope = %g, b = %g\n",__builtin___sprintf_chk (cmt, 0, __builtin_object_size (cmt, 2 > 1 ? 1 : 0), "%s: using thresh = %g, slope = %g, b = %g\n" , me, thresh, slope, b) |
| 1090 | me, thresh, slope, b)__builtin___sprintf_chk (cmt, 0, __builtin_object_size (cmt, 2 > 1 ? 1 : 0), "%s: using thresh = %g, slope = %g, b = %g\n" , me, thresh, slope, b); |
| 1091 | nrrdCommentAdd(nout, cmt); |
| 1092 | out = (float *)nout->data; |
| 1093 | for (I=0; I<(size_t)(sx*sy*sz); I++) { |
| 1094 | if (tenVerbose && !(I % (sx*sy))) { |
| 1095 | fprintf(stderr__stderrp, "%s: z = %d of %d\n", me, (int)(I/(sx*sy)), (int)sz-1); |
| 1096 | } |
| 1097 | chan[0] = nrrdFLookup[nin->type](nin->data, 0 + 7*I); |
| 1098 | chan[1] = nrrdFLookup[nin->type](nin->data, 1 + 7*I); |
| 1099 | chan[2] = nrrdFLookup[nin->type](nin->data, 2 + 7*I); |
| 1100 | chan[3] = nrrdFLookup[nin->type](nin->data, 3 + 7*I); |
| 1101 | chan[4] = nrrdFLookup[nin->type](nin->data, 4 + 7*I); |
| 1102 | chan[5] = nrrdFLookup[nin->type](nin->data, 5 + 7*I); |
| 1103 | chan[6] = nrrdFLookup[nin->type](nin->data, 6 + 7*I); |
| 1104 | calcten(tens, chan, thresh, slope, b); |
| 1105 | out[0 + 7*I] = tens[0]; |
| 1106 | out[1 + 7*I] = tens[1]; |
| 1107 | out[2 + 7*I] = tens[2]; |
| 1108 | out[3 + 7*I] = tens[3]; |
| 1109 | out[4 + 7*I] = tens[4]; |
| 1110 | out[5 + 7*I] = tens[5]; |
| 1111 | out[6 + 7*I] = tens[6]; |
| 1112 | } |
| 1113 | return 0; |
| 1114 | } |
| 1115 |