def batch(self, function, tech):
try:
filenames = utils.getFilenames('Load ' + tech + ' images',
[('Tiff file', '*.tif;*.tiff')],
self.folder)
nfiles = len(filenames)
function(filenames[0])
corrArray = np.zeros((nfiles, self.n[0], self.n[1]))
# Expand correlation array so it matches data shape
corrExp = np.empty((nfiles, self.initShape[0], self.initShape[1]),
dtype=np.single)
corrExp[:] = np.nan
ringsExp = np.empty((nfiles, self.initShape[0], self.initShape[1]),
dtype=np.single)
ringsExp[:] = np.nan
path = os.path.split(filenames[0])[0]
folder = os.path.split(path)[1]
self.folderStatus.setText('Processing folder ' + path)
print('Processing folder', path)
t0 = time.time()
# Make results directory if it doesn't exist
resultsDir = os.path.join(path, 'results')
if not os.path.exists(resultsDir):
os.makedirs(resultsDir)
resNames = [utils.insertFolder(p, 'results') for p in filenames]
for i in np.arange(nfiles):
print(os.path.split(filenames[i])[1])
self.fileStatus.setText(os.path.split(filenames[i])[1])
function(filenames[i])
self.ringFinder(False, batch=True)
corrArray[i] = self.localCorr
bound = (np.array(self.initShape) - self.crop).astype(np.int)
edge = bound - self.remanent
corrExp[i, self.crop:edge[0],
self.crop:edge[1]] = self.localCorrBig
# Save correlation values array
corrName = utils.insertSuffix(resNames[i], '_correlation')
tiff.imsave(corrName,
corrExp[i],
software='Gollum',
imagej=True,
resolution=(1000 / self.pxSize,
1000 / self.pxSize),
metadata={
'spacing': 1,
'unit': 'um'
})
# Saving ring images
ringsExp[corrExp < self.corrThres] = 0
ringsExp[corrExp >= self.corrThres] = 1
for i in np.arange(nfiles):
# Save correlation values array
ringName = utils.insertSuffix(resNames[i], '_rings')
tiff.imsave(ringName,
ringsExp[i],
software='Gollum',
imagej=True,
resolution=(1000 / self.pxSize,
1000 / self.pxSize),
metadata={
'spacing': 1,
'unit': 'um'
})
# save configuration file in the results folder
tools.saveConfig(self, os.path.join(resultsDir, 'config'))
# plot histogram of the correlation values
hrange = (np.min(np.nan_to_num(corrArray)),
np.max(np.nan_to_num(corrArray)))
y, x, _ = plt.hist(corrArray.flatten(), bins=20, range=hrange)
x = (x[1:] + x[:-1]) / 2
# Save data array as txt
corrArrayFlat = corrArray.flatten()
validCorr = corrArrayFlat[~np.isnan(corrArrayFlat)]
validArr = np.repeat(np.arange(nfiles), np.prod(self.n))
validArr = validArr[~np.isnan(corrArrayFlat)]
valuesTxt = os.path.join(resultsDir, folder + 'corr_values.txt')
corrByN = np.stack((validCorr, validArr), 1)
np.savetxt(valuesTxt, corrByN, fmt='%f\t%i')
groupedCorr = [
validCorr[np.where(validArr == i)] for i in np.arange(nfiles)
]
groupedCorr = [x for x in groupedCorr if len(x) > 0]
nfiles = len(groupedCorr)
meanCorrs = [np.mean(d) for d in groupedCorr]
validCorrRing = validCorr[np.where(validCorr > self.corrThres)]
validArrRing = validArr[np.where(validCorr > self.corrThres)]
groupedCorrRing = [
validCorrRing[np.where(validArrRing == i)]
for i in np.arange(nfiles)
]
ringFracs = [
len(groupedCorrRing[i]) / len(groupedCorr[i])
for i in np.arange(nfiles)
]
meanRingCorrs = np.array([np.mean(d) for d in groupedCorrRing])
meanRingCorrs = meanRingCorrs[~np.isnan(meanRingCorrs)]
n = corrArray.size - np.count_nonzero(np.isnan(corrArray))
nring = np.sum(validCorr > self.corrThres)
validRingCorr = validCorr[validCorr > self.corrThres]
ringFrac = nring / n
# Err estimation: stat err (binomial distribution, p=ringFrac)
statVar = ringFrac * (1 - ringFrac) / n
fracStd = math.sqrt(statVar)
statCorrVar = np.var(validCorr) / n
corrStd = math.sqrt(statCorrVar)
statRingCorrVar = np.var(validRingCorr) / nring
ringCorrStd = math.sqrt(statRingCorrVar)
# Plotting
plt.style.use('ggplot')
plt.figure(figsize=(10, 7.5))
plt.bar(x, y, align='center', width=(x[1] - x[0]), color="#3F5D7D")
plt.plot((self.corrThres, self.corrThres), (0, np.max(y)),
'--',
color='r',
linewidth=2)
text = ('Pearson coefficient threshold = {0:.2f} \n'
'n = {1}; nrings = {2} \n'
'PSS fraction = {3:.2f} $\pm$ {4:.2f} \n'
'mean coefficient = {5:.3f} $\pm$ {6:.3f}\n'
'mean ring coefficient = {7:.3f} $\pm$ {8:.3f}')
text = text.format(self.corrThres, n, nring, np.mean(ringFracs),
fracStd, np.mean(meanCorrs), corrStd,
np.mean(meanRingCorrs), ringCorrStd)
plt.text(0.75 * plt.axis()[1],
0.83 * plt.axis()[3],
text,
horizontalalignment='center',
verticalalignment='center',
bbox=dict(facecolor='white'),
fontsize=20)
plt.xlabel('Pearson correlation coefficient', fontsize=35)
plt.tick_params(axis='both', labelsize=25)
plt.grid()
plt.tight_layout()
plt.savefig(os.path.join(resultsDir, folder + 'corr_hist.pdf'),
dpi=300)
plt.savefig(os.path.join(resultsDir, folder + 'corr_hist.png'),
dpi=300)
plt.close()
folder = os.path.split(path)[1]
text = 'Folder ' + folder + ' done in {0:.0f} seconds'
print(text.format(time.time() - t0))
self.folderStatus.setText(text.format(time.time() - t0))
self.fileStatus.setText(' ')
except IndexError:
self.fileStatus.setText('No file selected!')
def updateConfig(self):
tools.saveConfig(self)