use of ij.gui.ExtendedGenericDialog in project GDSC-SMLM by aherbert.
the class BlinkEstimator method showDialog.
private boolean showDialog() {
ExtendedGenericDialog gd = new ExtendedGenericDialog(TITLE);
gd.addHelp(About.HELP_URL);
gd.addMessage("Compute the blinking rate by fitting counts to dark-time.\nSee Annibale et al (2011) PLos ONE 6, e22678.");
ResultsManager.addInput(gd, inputOption, InputSource.MEMORY);
gd.addNumericField("Max_dark_time (frames)", s_maxDarkTime, 0);
gd.addNumericField("Histogram_bins", histogramBins, 0);
gd.addCheckbox("Show_histogram", showHistogram);
gd.addSlider("Search_distance", 0.5, 5, s_searchDistance);
gd.addCheckbox("Relative_distance", s_relativeDistance);
gd.addSlider("Fitted_points", 4, 15, s_nFittedPoints);
gd.addSlider("Range_of_fitted_points", 0, 15, rangeFittedPoints);
gd.addCheckbox("Time_at_lower_bound", s_timeAtLowerBound);
//gd.addCheckbox("Fit_intercept", fitIntercept);
gd.showDialog();
if (gd.wasCanceled())
return false;
inputOption = gd.getNextChoice();
maxDarkTime = s_maxDarkTime = (int) gd.getNextNumber();
histogramBins = (int) gd.getNextNumber();
showHistogram = gd.getNextBoolean();
searchDistance = s_searchDistance = gd.getNextNumber();
relativeDistance = s_relativeDistance = gd.getNextBoolean();
nFittedPoints = s_nFittedPoints = (int) gd.getNextNumber();
rangeFittedPoints = (int) gd.getNextNumber();
timeAtLowerBound = s_timeAtLowerBound = gd.getNextBoolean();
// Check arguments
try {
Parameters.isAbove("Max dark time", maxDarkTime, 3);
Parameters.isAbove("Histogram bins", histogramBins, 1);
Parameters.isAboveZero("Search distance", searchDistance);
Parameters.isAbove("n-Fitted points", nFittedPoints, 3);
Parameters.isPositive("Range of fitted points", rangeFittedPoints);
} catch (IllegalArgumentException e) {
IJ.error(TITLE, e.getMessage());
return false;
}
return true;
}
use of ij.gui.ExtendedGenericDialog in project GDSC-SMLM by aherbert.
the class TraceMolecules method fitTraces.
private void fitTraces(MemoryPeakResults results, Trace[] traces) {
// Check if the original image is open and the fit configuration can be extracted
ImageSource source = results.getSource();
if (source == null)
return;
if (!source.open())
return;
FitEngineConfiguration config = (FitEngineConfiguration) XmlUtils.fromXML(results.getConfiguration());
if (config == null)
return;
// Show a dialog asking if the traces should be refit
ExtendedGenericDialog gd = new ExtendedGenericDialog(TITLE);
gd.addMessage("Do you want to fit the traces as a single peak using a combined image?");
gd.addCheckbox("Fit_closest_to_centroid", !fitOnlyCentroid);
gd.addSlider("Distance_threshold", 0.01, 3, distanceThreshold);
gd.addSlider("Expansion_factor", 1, 4.5, expansionFactor);
// Allow fitting settings to be adjusted
FitConfiguration fitConfig = config.getFitConfiguration();
gd.addMessage("--- Gaussian fitting ---");
String[] filterTypes = SettingsManager.getNames((Object[]) DataFilterType.values());
gd.addChoice("Spot_filter_type", filterTypes, filterTypes[config.getDataFilterType().ordinal()]);
String[] filterNames = SettingsManager.getNames((Object[]) DataFilter.values());
gd.addChoice("Spot_filter", filterNames, filterNames[config.getDataFilter(0).ordinal()]);
gd.addSlider("Smoothing", 0, 2.5, config.getSmooth(0));
gd.addSlider("Search_width", 0.5, 2.5, config.getSearch());
gd.addSlider("Border", 0.5, 2.5, config.getBorder());
gd.addSlider("Fitting_width", 2, 4.5, config.getFitting());
String[] solverNames = SettingsManager.getNames((Object[]) FitSolver.values());
gd.addChoice("Fit_solver", solverNames, solverNames[fitConfig.getFitSolver().ordinal()]);
String[] functionNames = SettingsManager.getNames((Object[]) FitFunction.values());
gd.addChoice("Fit_function", functionNames, functionNames[fitConfig.getFitFunction().ordinal()]);
String[] criteriaNames = SettingsManager.getNames((Object[]) FitCriteria.values());
gd.addChoice("Fit_criteria", criteriaNames, criteriaNames[fitConfig.getFitCriteria().ordinal()]);
gd.addNumericField("Significant_digits", fitConfig.getSignificantDigits(), 0);
gd.addNumericField("Coord_delta", fitConfig.getDelta(), 4);
gd.addNumericField("Lambda", fitConfig.getLambda(), 4);
gd.addNumericField("Max_iterations", fitConfig.getMaxIterations(), 0);
gd.addNumericField("Fail_limit", config.getFailuresLimit(), 0);
gd.addCheckbox("Include_neighbours", config.isIncludeNeighbours());
gd.addSlider("Neighbour_height", 0.01, 1, config.getNeighbourHeightThreshold());
gd.addSlider("Residuals_threshold", 0.01, 1, config.getResidualsThreshold());
//gd.addSlider("Duplicate_distance", 0, 1.5, fitConfig.getDuplicateDistance());
gd.addMessage("--- Peak filtering ---\nDiscard fits that shift; are too low; or expand/contract");
gd.addCheckbox("Smart_filter", fitConfig.isSmartFilter());
gd.addCheckbox("Disable_simple_filter", fitConfig.isDisableSimpleFilter());
gd.addSlider("Shift_factor", 0.01, 2, fitConfig.getCoordinateShiftFactor());
gd.addNumericField("Signal_strength", fitConfig.getSignalStrength(), 2);
gd.addNumericField("Min_photons", fitConfig.getMinPhotons(), 0);
gd.addSlider("Min_width_factor", 0, 0.99, fitConfig.getMinWidthFactor());
gd.addSlider("Width_factor", 1.01, 5, fitConfig.getWidthFactor());
gd.addNumericField("Precision", fitConfig.getPrecisionThreshold(), 2);
gd.addCheckbox("Debug_failures", debugFailures);
gd.showDialog();
if (!gd.wasOKed()) {
source.close();
return;
}
// Get parameters for the fit
fitOnlyCentroid = !gd.getNextBoolean();
distanceThreshold = (float) gd.getNextNumber();
expansionFactor = (float) gd.getNextNumber();
config.setDataFilterType(gd.getNextChoiceIndex());
config.setDataFilter(gd.getNextChoiceIndex(), Math.abs(gd.getNextNumber()), 0);
config.setSearch(gd.getNextNumber());
config.setBorder(gd.getNextNumber());
config.setFitting(gd.getNextNumber());
fitConfig.setFitSolver(gd.getNextChoiceIndex());
fitConfig.setFitFunction(gd.getNextChoiceIndex());
fitConfig.setFitCriteria(gd.getNextChoiceIndex());
fitConfig.setSignificantDigits((int) gd.getNextNumber());
fitConfig.setDelta(gd.getNextNumber());
fitConfig.setLambda(gd.getNextNumber());
fitConfig.setMaxIterations((int) gd.getNextNumber());
config.setFailuresLimit((int) gd.getNextNumber());
config.setIncludeNeighbours(gd.getNextBoolean());
config.setNeighbourHeightThreshold(gd.getNextNumber());
config.setResidualsThreshold(gd.getNextNumber());
fitConfig.setSmartFilter(gd.getNextBoolean());
fitConfig.setDisableSimpleFilter(gd.getNextBoolean());
fitConfig.setCoordinateShiftFactor(gd.getNextNumber());
fitConfig.setSignalStrength(gd.getNextNumber());
fitConfig.setMinPhotons(gd.getNextNumber());
fitConfig.setMinWidthFactor(gd.getNextNumber());
fitConfig.setWidthFactor(gd.getNextNumber());
fitConfig.setPrecisionThreshold(gd.getNextNumber());
// Check arguments
try {
Parameters.isAboveZero("Distance threshold", distanceThreshold);
Parameters.isAbove("Expansion factor", expansionFactor, 1);
Parameters.isAboveZero("Search_width", config.getSearch());
Parameters.isAboveZero("Fitting_width", config.getFitting());
Parameters.isAboveZero("Significant digits", fitConfig.getSignificantDigits());
Parameters.isAboveZero("Delta", fitConfig.getDelta());
Parameters.isAboveZero("Lambda", fitConfig.getLambda());
Parameters.isAboveZero("Max iterations", fitConfig.getMaxIterations());
Parameters.isPositive("Failures limit", config.getFailuresLimit());
Parameters.isPositive("Neighbour height threshold", config.getNeighbourHeightThreshold());
Parameters.isPositive("Residuals threshold", config.getResidualsThreshold());
Parameters.isPositive("Coordinate Shift factor", fitConfig.getCoordinateShiftFactor());
Parameters.isPositive("Signal strength", fitConfig.getSignalStrength());
Parameters.isPositive("Min photons", fitConfig.getMinPhotons());
Parameters.isPositive("Min width factor", fitConfig.getMinWidthFactor());
Parameters.isPositive("Width factor", fitConfig.getWidthFactor());
Parameters.isPositive("Precision threshold", fitConfig.getPrecisionThreshold());
} catch (IllegalArgumentException e) {
IJ.error(TITLE, e.getMessage());
source.close();
return;
}
debugFailures = gd.getNextBoolean();
if (!PeakFit.configureSmartFilter(globalSettings, filename))
return;
if (!PeakFit.configureDataFilter(globalSettings, filename, false))
return;
if (!PeakFit.configureFitSolver(globalSettings, filename, false))
return;
// Adjust settings for a single maxima
config.setIncludeNeighbours(false);
fitConfig.setDuplicateDistance(0);
// Create a fit engine
MemoryPeakResults refitResults = new MemoryPeakResults();
refitResults.copySettings(results);
refitResults.setName(results.getName() + " Trace Fit");
refitResults.setSortAfterEnd(true);
refitResults.begin();
// No border since we know where the peaks are and we must not miss them due to truncated searching
FitEngine engine = new FitEngine(config, refitResults, Prefs.getThreads(), FitQueue.BLOCKING);
// Either : Only fit the centroid
// or : Extract a bigger region, allowing all fits to run as normal and then
// find the correct spot using Euclidian distance.
// Set up the limits
final double stdDev = FastMath.max(fitConfig.getInitialPeakStdDev0(), fitConfig.getInitialPeakStdDev1());
float fitWidth = (float) (stdDev * config.getFitting() * ((fitOnlyCentroid) ? 1 : expansionFactor));
IJ.showStatus("Refitting traces ...");
List<JobItem> jobItems = new ArrayList<JobItem>(traces.length);
int singles = 0;
int fitted = 0;
for (int n = 0; n < traces.length; n++) {
Trace trace = traces[n];
if (n % 32 == 0)
IJ.showProgress(n, traces.length);
// Skip traces with one peak
if (trace.size() == 1) {
singles++;
// Use the synchronized method to avoid thread clashes with the FitEngine
refitResults.addSync(trace.getHead());
continue;
}
Rectangle bounds = new Rectangle();
double[] combinedNoise = new double[1];
float[] data = buildCombinedImage(source, trace, fitWidth, bounds, combinedNoise, false);
if (data == null)
continue;
// Fit the combined image
FitParameters params = new FitParameters();
params.noise = (float) combinedNoise[0];
float[] centre = trace.getCentroid();
if (fitOnlyCentroid) {
int newX = (int) Math.round(centre[0]) - bounds.x;
int newY = (int) Math.round(centre[1]) - bounds.y;
params.maxIndices = new int[] { newY * bounds.width + newX };
} else {
params.filter = new ArrayList<float[]>();
params.filter.add(new float[] { centre[0] - bounds.x, centre[1] - bounds.y });
params.distanceThreshold = distanceThreshold;
}
// This is not needed since the bounds are passed using the FitJob
//params.setOffset(new float[] { bounds.x, bounds.y });
int startT = trace.getHead().getFrame();
params.endT = trace.getTail().getFrame();
ParameterisedFitJob job = new ParameterisedFitJob(n, params, startT, data, bounds);
jobItems.add(new JobItem(job, trace, centre));
engine.run(job);
fitted++;
}
engine.end(false);
IJ.showStatus("");
IJ.showProgress(1);
// Check the success ...
FitStatus[] values = FitStatus.values();
int[] statusCount = new int[values.length + 1];
ArrayList<String> names = new ArrayList<String>(Arrays.asList(SettingsManager.getNames((Object[]) values)));
names.add(String.format("No maxima within %.2f of centroid", distanceThreshold));
int separated = 0;
int success = 0;
final int debugLimit = 3;
for (JobItem jobItem : jobItems) {
int id = jobItem.getId();
ParameterisedFitJob job = jobItem.job;
Trace trace = jobItem.trace;
int[] indices = job.getIndices();
FitResult fitResult = null;
int status;
if (indices.length < 1) {
status = values.length;
} else if (indices.length > 1) {
// Choose the first OK result. This is all that matters for the success reporting
for (int n = 0; n < indices.length; n++) {
if (job.getFitResult(n).getStatus() == FitStatus.OK) {
fitResult = job.getFitResult(n);
break;
}
}
// Otherwise use the closest failure.
if (fitResult == null) {
final float[] centre = traces[id].getCentroid();
double minD = Double.POSITIVE_INFINITY;
for (int n = 0; n < indices.length; n++) {
// Since the fit has failed we use the initial parameters
final double[] params = job.getFitResult(n).getInitialParameters();
final double dx = params[Gaussian2DFunction.X_POSITION] - centre[0];
final double dy = params[Gaussian2DFunction.Y_POSITION] - centre[1];
final double d = dx * dx + dy * dy;
if (minD > d) {
minD = d;
fitResult = job.getFitResult(n);
}
}
}
status = fitResult.getStatus().ordinal();
} else {
fitResult = job.getFitResult(0);
status = fitResult.getStatus().ordinal();
}
// All jobs have only one peak
statusCount[status]++;
// Debug why any fits failed
if (fitResult == null || fitResult.getStatus() != FitStatus.OK) {
refitResults.addAll(trace.getPoints());
separated += trace.size();
if (debugFailures) {
FitStatus s = (fitResult == null) ? FitStatus.UNKNOWN : fitResult.getStatus();
// Only display the first n per category to limit the number of images
double[] noise = new double[1];
if (statusCount[status] <= debugLimit) {
Rectangle bounds = new Rectangle();
buildCombinedImage(source, trace, fitWidth, bounds, noise, true);
float[] centre = trace.getCentroid();
Utils.display(String.format("Trace %d (n=%d) : x=%f,y=%f", id, trace.size(), centre[0], centre[1]), slices);
switch(s) {
case INSUFFICIENT_PRECISION:
float precision = (Float) fitResult.getStatusData();
IJ.log(String.format("Trace %d (n=%d) : %s = %f", id, trace.size(), names.get(status), precision));
break;
case INSUFFICIENT_SIGNAL:
if (noise[0] == 0)
noise[0] = getCombinedNoise(trace);
float snr = (Float) fitResult.getStatusData();
IJ.log(String.format("Trace %d (n=%d) : %s = %f (noise=%.2f)", id, trace.size(), names.get(status), snr, noise[0]));
break;
case COORDINATES_MOVED:
case OUTSIDE_FIT_REGION:
case WIDTH_DIVERGED:
float[] shift = (float[]) fitResult.getStatusData();
IJ.log(String.format("Trace %d (n=%d) : %s = %.3f,%.3f", id, trace.size(), names.get(status), shift[0], shift[1]));
break;
default:
IJ.log(String.format("Trace %d (n=%d) : %s", id, trace.size(), names.get(status)));
break;
}
}
}
} else {
success++;
if (debugFailures) {
// Only display the first n per category to limit the number of images
double[] noise = new double[1];
if (statusCount[status] <= debugLimit) {
Rectangle bounds = new Rectangle();
buildCombinedImage(source, trace, fitWidth, bounds, noise, true);
float[] centre = trace.getCentroid();
Utils.display(String.format("Trace %d (n=%d) : x=%f,y=%f", id, trace.size(), centre[0], centre[1]), slices);
}
}
}
}
IJ.log(String.format("Trace fitting : %d singles : %d / %d fitted : %d separated", singles, success, fitted, separated));
if (separated > 0) {
IJ.log("Reasons for fit failure :");
// Start at i=1 to skip FitStatus.OK
for (int i = 1; i < statusCount.length; i++) {
if (statusCount[i] != 0)
IJ.log(" " + names.get(i) + " = " + statusCount[i]);
}
}
refitResults.end();
MemoryPeakResults.addResults(refitResults);
source.close();
}
use of ij.gui.ExtendedGenericDialog in project GDSC-SMLM by aherbert.
the class TraceMolecules method showDialog.
private boolean showDialog() {
TITLE = outputName + " Molecules";
ExtendedGenericDialog gd = new ExtendedGenericDialog(TITLE);
gd.addHelp(About.HELP_URL);
ResultsManager.addInput(gd, inputOption, InputSource.MEMORY);
globalSettings = SettingsManager.loadSettings();
settings = globalSettings.getClusteringSettings();
gd.addNumericField("Distance_Threshold (nm)", settings.distanceThreshold, 2);
gd.addNumericField("Distance_Exclusion (nm)", settings.distanceExclusion, 2);
gd.addNumericField("Time_Threshold", settings.getTimeThreshold(), 2);
String[] timeUnits = SettingsManager.getNames((Object[]) ClusteringSettings.TimeUnit.values());
gd.addChoice("Time_unit", timeUnits, timeUnits[settings.getTimeUnit().ordinal()]);
String[] traceModes = SettingsManager.getNames((Object[]) TraceManager.TraceMode.values());
gd.addChoice("Trace_mode", traceModes, traceModes[settings.getTraceMode().ordinal()]);
gd.addNumericField("Pulse_interval (frames)", settings.pulseInterval, 0);
gd.addNumericField("Pulse_window (frames)", settings.pulseWindow, 0);
gd.addCheckbox("Split_pulses", settings.splitPulses);
gd.addCheckbox("Optimise", settings.optimise);
gd.addCheckbox("Save_traces", settings.saveTraces);
gd.addCheckbox("Show_histograms", settings.showHistograms);
gd.addCheckbox("Save_trace_data", settings.saveTraceData);
gd.addCheckbox("Refit_option", settings.refitOption);
if (altKeyDown) {
gd.addCheckbox("Debug", inputDebugMode);
}
gd.showDialog();
if (gd.wasCanceled() || !readDialog(gd))
return false;
// Update the settings
SettingsManager.saveSettings(globalSettings);
// Load the results
results = ResultsManager.loadInputResults(inputOption, true);
if (results == null || results.size() == 0) {
IJ.error(TITLE, "No results could be loaded");
IJ.showStatus("");
return false;
}
// Store exposure time in seconds
exposureTime = results.getCalibration().getExposureTime() / 1000;
return true;
}
use of ij.gui.ExtendedGenericDialog in project GDSC-SMLM by aherbert.
the class TraceMolecules method readDialog.
private boolean readDialog(ExtendedGenericDialog gd) {
inputOption = ResultsManager.getInputSource(gd);
settings.distanceThreshold = gd.getNextNumber();
settings.distanceExclusion = gd.getNextNumber();
settings.setTimeThreshold(gd.getNextNumber());
settings.setTimeUnit(gd.getNextChoiceIndex());
settings.setTraceMode(gd.getNextChoiceIndex());
settings.pulseInterval = (int) gd.getNextNumber();
settings.pulseWindow = (int) gd.getNextNumber();
settings.splitPulses = gd.getNextBoolean();
settings.optimise = gd.getNextBoolean();
settings.saveTraces = gd.getNextBoolean();
settings.showHistograms = gd.getNextBoolean();
settings.saveTraceData = gd.getNextBoolean();
settings.refitOption = gd.getNextBoolean();
if (altKeyDown) {
debugMode = inputDebugMode = gd.getNextBoolean();
}
if (gd.invalidNumber())
return false;
if (settings.showHistograms) {
gd = new ExtendedGenericDialog(TITLE);
gd.addMessage("Select the histograms to display");
gd.addCheckbox("Remove_outliers", settings.removeOutliers);
gd.addNumericField("Histogram_bins", settings.histogramBins, 0);
for (int i = 0; i < displayHistograms.length; i++) gd.addCheckbox(NAMES[i].replace(' ', '_'), displayHistograms[i]);
gd.showDialog();
if (gd.wasCanceled())
return false;
settings.removeOutliers = gd.getNextBoolean();
settings.histogramBins = (int) Math.abs(gd.getNextNumber());
for (int i = 0; i < displayHistograms.length; i++) displayHistograms[i] = gd.getNextBoolean();
}
// Check arguments
try {
Parameters.isAboveZero("Distance threshold", settings.distanceThreshold);
Parameters.isAboveZero("Time threshold", settings.getTimeThreshold());
Parameters.isPositive("Pulse interval", settings.pulseInterval);
Parameters.isPositive("Pulse window", settings.pulseWindow);
Parameters.isAboveZero("Histogram bins", settings.histogramBins);
} catch (IllegalArgumentException e) {
IJ.error(TITLE, e.getMessage());
return false;
}
return true;
}
use of ij.gui.ExtendedGenericDialog in project GDSC-SMLM by aherbert.
the class CalibrateResults method showInputDialog.
private boolean showInputDialog() {
int size = MemoryPeakResults.countMemorySize();
if (size == 0) {
IJ.error(TITLE, "There are no fitting results in memory");
return false;
}
ExtendedGenericDialog gd = new ExtendedGenericDialog(TITLE);
gd.addHelp(About.HELP_URL);
gd.addMessage("Select results to calibrate");
ResultsManager.addInput(gd, inputOption, InputSource.MEMORY);
gd.showDialog();
if (gd.wasCanceled())
return false;
inputOption = ResultsManager.getInputSource(gd);
return true;
}
Aggregations