// plug in finalization function
DLLExport finalSVM(studyParams &vdb, void *&userData)
{
if (userData != NULL)
{
SVMProcessing *svmProcessingVar = (SVMProcessing *) userData;
svmProcessingVar->cleanUp();
if (svmProcessingVar->hasPredicted)
{
char confusionMatrixFile[500];
sprintf(confusionMatrixFile, "%s%s%s.txt", svmProcessingVar->svmDir, "confusionMatrix", vdb.trainFeatureSuffix);
svmProcessingVar->accuracyResults.saveMatrixReport(confusionMatrixFile);
if (fileExists(svmProcessingVar->projectionsFilename))
{
stringstream CmdLn;
char pngFile[BUFF_SIZE];
// generating the projections graph png of the actual run
changeFileExt(svmProcessingVar->projectionsFilename, ".png", pngFile);
fprintf(stderr, "Generating svm projection graphics of the current run performance\n");
CmdLn << "fsl_tsplot -i " << svmProcessingVar->projectionsFilename << " -t \"SVM projections\" -u 1 --start=1 --finish=1 -a Projections -w 640 -h 144 -o " << pngFile;
fsl_tsplot((char *)CmdLn.str().c_str());
}
}
delete svmProcessingVar;
userData = NULL;
}
return 0;
}
// function that creates a roi map from glm and a previously defined roi in mni
void emotionRoiProcessing::createROIVolume(studyParams &vdb)
{
// Bring MNI Mask to Subject Space
char compositeFile[500], outputFile[500], name[500], regionExtractMapFile[500], roiVolumeFile[500], meanFile[500];
char pngFile[500];
double correlationExtractPercent;
stringstream CmdLn;
char prefix[30] = "_RFI2";
vector<vector<double>> timeseries;
// generating the means file
volume<float> v;
// preparing timeseries variable
timeseries.resize(meanCalculation.roiCount());
for (int i = 0; i < timeseries.size(); i++)
timeseries[i].resize(vdb.interval.maxIndex());
// calculating the means
for (int i = 0; i < vdb.interval.maxIndex(); i++)
{
loadVolume(vdb, v, i+1);
meanCalculation.calculateMeans(v);
for (int j = 0; j < meanCalculation.roiCount(); j++)
timeseries[j][i] = meanCalculation.roiMean(j);
}
// z normalized
for (int j = 0; j < timeseries.size(); j++) znormalise(timeseries[j]);
// saving the result to file
sprintf(meanFile, "%s%s_%s%s.txt", vdb.outputDir, vdb.subject, "means", vdb.trainFeatureSuffix);
fstream Output(meanFile, fstream::in | fstream::out | fstream::trunc);
for (int i = 0; i < vdb.interval.maxIndex(); i++)
{
for (int j = 0; j < timeseries.size(); j++)
Output << timeseries[j][i] << '\t';
Output << '\n';
}
Output.close();
// generating the roi means graph png
fprintf(stderr, "Generating roi means graphic\n");
changeFileExt(meanFile, ".png", pngFile);
CmdLn << "fsl_tsplot -i " << meanFile << " -t \"z-normalised roi means plot\" -u 1 --start=1 --finish=" << meanCalculation.roiCount() << " -a ";
vector<int> roiValues;
meanCalculation.getRoiValues(roiValues);
// Building the labels with the intensities of the roi volume file
int counter=0;
CmdLn << '\"';
for (int t = 0; t < roiValues.size(); t++)
{
CmdLn << "Intensity " << roiValues[t];
counter++;
if (counter < meanCalculation.roiCount())
CmdLn << ',';
}
// completing the command
CmdLn << '\"';
CmdLn << " -w 640 -h 144 -o " << pngFile;
fsl_tsplot((char *)CmdLn.str().c_str());
extractFileName(vdb.mniMask, name);
for (int t = 0; t<strlen(name); t++)
if (name[t] == '.') name[t] = '_';
sprintf(outputFile, "%s%s%s.nii", vdb.inputDir, name, vdb.trainFeatureSuffix);
MniToSubject(vdb.maskFile, vdb.mniMask, vdb.mniTemplate, outputFile, prefix);
// saving a composite file just to assure the side of the roi volume
sprintf(compositeFile, "%s%s%s.nii", vdb.inputDir, "compositeFile", vdb.trainFeatureSuffix);
uniteVolumes(vdb.rfiFile, outputFile, compositeFile);
// read Region extract map file and percentage
correlationExtractPercent = vdb.readedIni.GetDoubleValue("FRIEND", "PercentageOfBestVoxelsPerROI") / 100.0;
strcpy(regionExtractMapFile, vdb.readedIni.GetValue("FRIEND", "ROIExtractionMapFile"));
RegionExtraction extractor;
// reading the mappings
std::map<int, int> mappings;
extractor.readMappings(regionExtractMapFile, mappings);
sprintf(roiVolumeFile, "%s%s%s%s", vdb.outputDir, "ROIsMap", vdb.trainFeatureSuffix, ".nii");
// Execute segmentation
extractor.regionsExtraction(outputFile, vdb.glmTOutput, vdb.featuresAllTrainSuffix, roiVolumeFile, mappings, correlationExtractPercent);
// calculate initial target values for positive and negative conditions
if (fileExists(roiVolumeFile))
{
meanCalculation.loadReference(roiVolumeFile);
positiveIndex = meanCalculation.roiIndex(3); // 3 in the intensity value of positive emotion ROI
negativeIndex = meanCalculation.roiIndex(1); // 1 in the intensity value of negative emotion ROI
for (int i=1; i <= vdb.interval.maxIndex(); i++)
{
float classnum, projection;
processVolume(vdb, i, classnum, projection);
}
char negativeCurveFile[BUFF_SIZE], positiveCurveFile[BUFF_SIZE];
sprintf(negativeCurveFile, "%s%c%s%s.txt", vdb.outputDir, PATHSEPCHAR, "negative_curve", vdb.trainFeatureSuffix);
sprintf(positiveCurveFile, "%s%c%s%s.txt", vdb.outputDir, PATHSEPCHAR, "positive_curve", vdb.trainFeatureSuffix);
positiveActivationLevel.saveCurves(positiveCurveFile);
negativeActivationLevel.saveCurves(negativeCurveFile);
vdb.readedIni.SetDoubleValue("FRIEND", "NegativeActivationLevel", negativeActivationLevel.mean);
vdb.readedIni.SetDoubleValue("FRIEND", "PositiveActivationLevel", positiveActivationLevel.mean);
vdb.readedIni.SaveFile(vdb.configFileNameRead);
}
}
