def run(self):
self.signals.start.emit()
localCorr = np.zeros(len(self.blocksInput))
# Single-core code
for i in np.arange(len(self.blocksInput)):
rings = False
block = self.blocksInput[i]
blockS = self.blocksInputS[i]
mask = self.blocksMask[i]
# Block may be excluded from the analysis for two reasons.
# Firstly, because the intensity for all its pixels may be
# too low. Secondly, because the part of the block that
# belongs toa neuron may be below an arbitrary 30% of the
# block. We apply intensity threshold to smoothed data so we
# don't catch tiny bright spots outside neurons
neuronFrac = 1 - np.sum(mask) / np.size(mask)
thres = self.meanS + self.intThr * self.stdS
if np.any(blockS > thres) and neuronFrac > 0.25:
output = tools.corrMethod(block, mask, *self.cArgs)
angle, corrTheta, corrMax, theta, phase, rings = output
# Store results
localCorr[i] = corrMax
else:
localCorr[i] = np.nan
localCorr = localCorr.reshape(*self.n)
self.signals.done.emit(localCorr)
def ringFinder(self, show=True, batch=False):
"""RingFinder handles the input data, and then evaluates every subimg
using the given algorithm which decides if there are rings or not.
Subsequently gives the output data and plots it"""
if self.corrButton.isChecked() or batch:
self.corrResult.clear()
self.ringResult.clear()
# for each subimg, we apply the correlation method for ring finding
intThr = np.float(self.intThresEdit.text())
minLen = np.float(self.lineLengthEdit.text()) / self.pxSize
thetaStep = np.float(self.thetaStepEdit.text())
deltaTh = np.float(self.deltaThEdit.text())
wvlen = np.float(self.wvlenEdit.text()) / self.pxSize
sinPow = np.float(self.sinPowerEdit.text())
cArgs = minLen, thetaStep, deltaTh, wvlen, sinPow
# Single-core code
self.localCorr = np.zeros(len(self.blocksInput))
thres = self.meanS + intThr * self.stdS
if not (self.testData):
thres = thres * np.ones(self.blocksInput.shape)
for i in np.arange(len(self.blocksInput)):
block = self.blocksInput[i]
blockS = self.blocksInputS[i]
mask = self.blocksMask[i]
# Block may be excluded from the analysis for two reasons.
# Firstly, because the intensity for all its pixels may be
# too low. Secondly, because the part of the block that
# belongs toa neuron may be below an arbitrary 20% of the
# block. We apply intensity threshold to smoothed data so we
# don't catch tiny bright spots outside neurons
neuronFrac = 1 - np.sum(mask) / np.size(mask)
areaThres = 0.01 * float(self.minAreaEdit.text())
if np.any(blockS > thres[i]) and neuronFrac > areaThres:
output = tools.corrMethod(block, mask, *cArgs)
angle, corrTheta, corrMax, theta, phase = output
# Store results
self.localCorr[i] = corrMax
else:
self.localCorr[i] = np.nan
self.localCorr = self.localCorr.reshape(*self.n)
self.updateGUI(self.localCorr)
else:
self.corrResult.clear()
self.ringResult.clear()
def corrMethodGUI(self):
self.pCorr.clear()
# We apply intensity threshold to smoothed data so we don't catch
# tiny bright spots outside neurons
thr = np.float(self.main.intThresEdit.text())
if np.any(self.selectedS > self.selectedMean + thr * self.selectedStd):
self.getDirection()
# we apply the correlation method for ring finding for the
# selected subimg
minLen = np.float(self.main.lineLengthEdit.text()) / self.pxSize
thStep = np.float(self.main.thetaStepEdit.text())
deltaTh = np.float(self.main.deltaThEdit.text())
wvlen = np.float(self.main.wvlenEdit.text()) / self.pxSize
sinPow = np.float(self.main.sinPowerEdit.text())
args = [self.selectedMask, minLen, thStep, deltaTh, wvlen, sinPow]
output = tools.corrMethod(self.selected, *args, developer=True)
self.th0, corrTheta, corrMax, thetaMax, phaseMax = output
if np.all([self.th0, corrMax]) is not None:
self.bestAxon = simAxon(imSize=self.subImgSize,
wvlen=wvlen,
theta=thetaMax,
phase=phaseMax,
b=sinPow).data
self.bestAxon = np.ma.array(self.bestAxon,
mask=self.selectedMask,
fill_value=0)
self.img1.setImage(self.bestAxon.filled(0))
self.img2.setImage(self.selected)
shape = self.selected.shape
self.vb4.setLimits(xMin=-0.05 * shape[0],
xMax=1.05 * shape[0],
yMin=-0.05 * shape[1],
yMax=1.05 * shape[1],
minXRange=4,
minYRange=4)
self.vb4.setRange(xRange=(0, shape[0]), yRange=(0, shape[1]))
# plot the threshold of correlation chosen by the user
# phase steps are set to 20, TO DO: explore this parameter
theta = np.arange(np.min([self.th0 - deltaTh, 0]), 180, thStep)
pen1 = pg.mkPen(color=(0, 255, 100),
width=2,
style=QtCore.Qt.SolidLine,
antialias=True)
self.pCorr.plot(theta, corrTheta, pen=pen1)
# plot the area within deltaTh from the found direction
if self.th0 is not None:
thArea = np.arange(self.th0 - deltaTh, self.th0 + deltaTh)
if self.th0 < 0:
thArea += 180
gap = 0.05 * (np.max(corrTheta) - np.min(corrTheta))
brushMax = np.max(corrTheta) * np.ones(len(thArea)) + gap
brushMin = np.min(corrTheta) - gap
self.pCorr.plot(thArea,
brushMax,
fillLevel=brushMin,
fillBrush=(50, 50, 200, 100),
pen=None)
corrThres = np.float(self.main.corrThresEdit.text())
rings = corrMax > corrThres
if rings and np.abs(self.th0 - thetaMax) <= deltaTh:
self.main.resultLabel.setText('<strong>MY PRECIOUS!<\strong>')
else:
if rings:
print('Correlation maximum outside direction theta range')
self.main.resultLabel.setText('<strong>No rings<\strong>')
else:
print('Data below intensity threshold')
self.main.resultLabel.setText('<strong>No rings<\strong>')