def getpoints(input_dir, num_points=5, num_pixels=10.0):
"""
Evaluate spline at discrete points that will be compared to establish
continuous time trajectories.
args:
df (DataFrame): contains requested coordinates and trajectories
kwargs:
num_points (int): number of points to compare
num_pixels (float): number of pixels along midline to compare
"""
# Read the midline and any previous trajectories in from file
df = read.makedataframe(input_dir, ('MidSpline', 'Length', 'Trace'))
df = df.reindex(sorted(df.index))
if 'Trace' not in df:
df['Trace'] = [v for v in df.index]
# Evaluate the midline at discrete points starting at the old pole
u0 = np.linspace(0., 1., num_points)
vs = []
for i in df.index:
u = u0 * min(1, num_pixels / df['Length'][i])
try:
v = splev(u, df['MidSpline'][i])
except:
# traceback.print_exc()
v = [np.ones(num_points) * np.nan] * 2
vs.append(v)
df['Points'] = vs
return df.ix[:, ('Points', 'Trace')]
def get_fft_coefs(input_dir, num_points=500, num_coefs=20):
"""
Calculate the Fourier shape descriptors from the saved cell contours.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
df = read.makedataframe(input_dir, ('EdgeSpline', ))
df['FourierCoef'] = None
# Evaluate each frame independently
u = np.arange(0., 1., 1. / num_points)
for i, v in enumerate(df.index):
tck = df['EdgeSpline'].ix[v]
try:
xs = splineprops(u, tck, 'Coordinates')['Coordinates']
d = fftcoefs(xs, n=num_coefs)
except Exception:
d = {}
df['FourierCoef'].ix[v] = d
# Save new Series to file
df['FourierCoef'].to_pickle(os.path.join(input_dir, 'FourierCoef.pickle'))
def get_width_profiles(input_dir, num_points=500):
"""
Calculate the width profiles from the saved cell midlines.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
old_keys = ('Area', 'EdgeSpline', 'Length', 'MidSpline', 'Perimeter',
'StalkedPole', 'SwarmerPole', 'Trace')
df = read.makedataframe(input_dir, old_keys)
# Evaluate each frame independently
for i, v in enumerate(df.index):
d = get_width_profile_new(df.ix[v], num_points)
if i == 0:
new_keys = d.keys()
for k in new_keys:
df[k] = None
for k in new_keys:
df[k].ix[v] = d[k]
# Save each Series to file
for k in new_keys:
df[k].to_pickle(os.path.join(input_dir, k + '.pickle'))
def get_fft_coefs(input_dir, num_points=500, num_coefs=20):
"""
Calculate the Fourier shape descriptors from the saved cell contours.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
df = read.makedataframe(input_dir, ("EdgeSpline",))
df["FourierCoef"] = None
# Evaluate each frame independently
u = np.arange(0.0, 1.0, 1.0 / num_points)
for i, v in enumerate(df.index):
tck = df["EdgeSpline"].ix[v]
try:
xs = splineprops(u, tck, "Coordinates")["Coordinates"]
d = fftcoefs(xs, n=num_coefs)
except Exception:
d = {}
df["FourierCoef"].ix[v] = d
# Save new Series to file
df["FourierCoef"].to_pickle(os.path.join(input_dir, "FourierCoef.pickle"))
def get_width_profiles(input_dir, num_points=500):
"""
Calculate the width profiles from the saved cell midlines.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
old_keys = ("Area", "EdgeSpline", "Length", "MidSpline", "Perimeter", "StalkedPole", "SwarmerPole", "Trace")
df = read.makedataframe(input_dir, old_keys)
# Evaluate each frame independently
for i, v in enumerate(df.index):
d = get_width_profile_new(df.ix[v], num_points)
if i == 0:
new_keys = d.keys()
for k in new_keys:
df[k] = None
for k in new_keys:
df[k].ix[v] = d[k]
# Save each Series to file
for k in new_keys:
df[k].to_pickle(os.path.join(input_dir, k + ".pickle"))
def getpoints(input_dir, num_points=5, num_pixels=10.0):
"""
Evaluate spline at discrete points that will be compared to establish
continuous time trajectories.
args:
df (DataFrame): contains requested coordinates and trajectories
kwargs:
num_points (int): number of points to compare
num_pixels (float): number of pixels along midline to compare
"""
# Read the midline and any previous trajectories in from file
df = read.makedataframe(input_dir, ('MidSpline', 'Length', 'Trace'))
df = df.reindex(sorted(df.index))
if 'Trace' not in df:
df['Trace'] = [v for v in df.index]
# Evaluate the midline at discrete points starting at the old pole
u0 = np.linspace(0., 1., num_points)
vs = []
for i in df.index:
u = u0 * min(1, num_pixels / df['Length'][i])
try:
v = splev(u, df['MidSpline'][i])
except:
# traceback.print_exc()
v = [np.ones(num_points) * np.nan] * 2
vs.append(v)
df['Points'] = vs
return df.ix[:, ('Points', 'Trace')]
def smooth_widths(input_dir, sigma=2, cutoff=0.02):
"""
Smooth the widths and recalculate the shape parameters.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
df = read.makedataframe(input_dir,
('Area', 'EdgeSpline', 'Length', 'MidSpline',
'StalkedPole', 'SwarmerPole', 'Widths'))
# Save new and overwrite existing shape variables
new_keys = ('AreaStalked', 'LengthStalkedMaxPole', 'LengthStalkedMin',
'LengthStalkedMaxCenter', 'LengthMin',
'LengthSwarmerMaxCenter', 'LengthSwarmerMin',
'LengthSwarmerMaxPole', 'WidthStalkedMaxPole',
'WidthStalkedMin', 'WidthStalkedMaxCenter', 'WidthMin',
'WidthSwarmerMaxCenter', 'WidthSwarmerMin',
'WidthSwarmerMaxPole', 'WidthStalkedMax', 'WidthSwarmerMax',
'WidthMax', 'LengthStalkedMax', 'LengthSwarmerMax',
'LengthMax', 'WidthsSmoothed')
for k in new_keys:
df[k] = None
# Evaluate each frame independently
for v in df.index:
# Assign default values
y = np.ndarray(0)
ast = np.nan
lstmaxp = np.nan
lstmin = np.nan
lstmaxc = np.nan
lst = np.nan
lswmaxc = np.nan
lswmin = np.nan
lswmaxp = np.nan
wstmaxp = np.nan
wstmin = np.nan
wstmaxc = np.nan
wmin = np.nan
wswmaxc = np.nan
wswmin = np.nan
wswmaxp = np.nan
wstmax = np.nan
wswmax = np.nan
wmax = np.nan
lstmax = np.nan
lswmax = np.nan
lmax = np.nan
try:
s = df.ix[v]
# Load variables from file
a = s['Area']
l = s['Length']
ps = (s['StalkedPole'], s['SwarmerPole'])
w = s['Widths']
tcke = s['EdgeSpline']
tckm = s['MidSpline']
# Reconstruct cell contour and midline
n = len(w)
u = np.linspace(0., 1., n)
es = np.asarray(zip(*splev(u, tcke)))
ms = np.asarray(zip(*splev(u, tckm)))
dms = np.asarray(zip(*splev(u, tckm, der=1)))
# Smooth the width profile
y = gaussian_filter1d(w, sigma=sigma)
# Truncate the width profile near the poles
#y = y[m:-m]
# Find all relative minima and maxima
m = int(n * cutoff)
imin = np.hstack(argrelmin(y, order=m))
imax = np.hstack(argrelmax(y, order=m))
if np.any(imin):
# Find the smallest (primary) minimum
jmin = imin[np.argmin(y[imin])]
# Assign the minimum width
wmin = y[jmin]
# Find the normal line that split the cell at the minimum width
esmin = get_divn_plane(es, ms[jmin], dms[jmin])
# Split into stalked and swarmer parts
ast, _ = split_areas(u, tcke, esmin, es, ps, a)
lst = get_length(u[:jmin + 1], tckm)
# Sort the other minima to the stalked or swarmer part
istmin = imin[imin < jmin]
iswmin = imin[imin > jmin]
# Sort the maxima to the stalked or swarmer part
istmax = imax[imax < jmin]
iswmax = imax[imax > jmin]
if np.any(istmin):
# Find the smallest of the stalked minima
jstmin = istmin[np.argmin(y[istmin])]
# Assign the minimum stalked width
wstmin = y[jstmin]
# Find the length at the minimum stalked width
lstmin = get_length(u[:jstmin + 1], tckm)
# Sort the stalked maxima to the pole or center
istmaxp = istmax[istmax < jstmin]
istmaxc = istmax[istmax > jstmin]
# Find the largest of the stalked maxima
jstmaxp = istmaxp[np.argmax(y[istmaxp])]
jstmaxc = istmaxc[np.argmax(y[istmaxc])]
# Assign the stalked pole width and length maxima
wstmaxp = y[jstmaxp]
lstmaxp = get_length(u[:jstmaxp + 1], tckm)
# Assign the stalked center width and length maxima
wstmaxc = y[jstmaxc]
lstmaxc = get_length(u[:jstmaxc + 1], tckm)
else:
# Find the largest of the stalked maxima
jstmax = istmax[np.argmax(y[istmax])]
# Assign the stalked pole width and length maxima
wstmax = y[jstmax]
lstmax = get_length(u[:jstmax + 1], tckm)
if np.any(iswmin):
# Find the smallest of the swarmer minima
jswmin = iswmin[np.argmin(y[iswmin])]
# Assign the minimum swarmer width
wswmin = y[jswmin]
# Find the length at the minimum swarmer width
lswmin = get_length(u[:jswmin + 1], tckm)
# Sort the swarmer maxima to the pole or center
iswmaxp = iswmax[iswmax > jswmin]
iswmaxc = iswmax[iswmax < jswmin]
# Find the largest of the swarmer maxima
jswmaxp = iswmaxp[np.argmax(y[iswmaxp])]
jswmaxc = iswmaxc[np.argmax(y[iswmaxc])]
# Assign the swarmer pole width and length maxima
wswmaxp = y[jswmaxp]
lswmaxp = get_length(u[:jswmaxp + 1], tckm)
# Assign the swarmer center width and length maxima
wswmaxc = y[jswmaxc]
lswmaxc = get_length(u[:jswmaxc + 1], tckm)
else:
# Find the largest of the swarmer maxima
jswmax = iswmax[np.argmax(y[iswmax])]
# Assign the swarmer pole width and length maxima
wswmax = y[jswmax]
lswmax = get_length(u[:jswmax + 1], tckm)
else:
# Find the largest of the maxima
jmax = imax[np.argmax(y[imax])]
# Assign the width and length maxima
wmax = y[jmax]
lmax = get_length(u[:jmax + 1], tckm)
except Exception as e:
pass
s['WidthsSmoothed'] = y
s['AreaStalked'] = ast
s['LengthStalkedMaxPole'] = lstmaxp
s['LengthStalkedMin'] = lstmin
s['LengthStalkedMaxCenter'] = lstmaxc
s['LengthMin'] = lst
s['LengthSwarmerMaxCenter'] = lswmaxc
s['LengthSwarmerMin'] = lswmin
s['LengthSwarmerMaxPole'] = lswmaxp
s['WidthStalkedMaxPole'] = wstmaxp
s['WidthStalkedMin'] = wstmin
s['WidthStalkedMaxCenter'] = wstmaxc
s['WidthMin'] = wmin
s['WidthSwarmerMaxCenter'] = wswmaxc
s['WidthSwarmerMin'] = wswmin
s['WidthSwarmerMaxPole'] = wswmaxp
s['WidthStalkedMax'] = wstmax
s['WidthSwarmerMax'] = wswmax
s['WidthMax'] = wmax
s['LengthStalkedMax'] = lstmax
s['LengthSwarmerMax'] = lswmax
s['LengthMax'] = lmax
df.ix[v] = s
# Save new Series to file
for k in new_keys:
df[k].to_pickle(os.path.join(input_dir, k + '.pickle'))
def smooth_widths(input_dir, sigma=2, cutoff=0.02):
"""
Smooth the widths and recalculate the shape parameters.
args:
df (DataFrame): contains requested coordinates and trajectories
"""
# Read the necessary variables in from file
df = read.makedataframe(
input_dir, ("Area", "EdgeSpline", "Length", "MidSpline", "StalkedPole", "SwarmerPole", "Widths")
)
# Save new and overwrite existing shape variables
new_keys = (
"AreaStalked",
"LengthStalkedMaxPole",
"LengthStalkedMin",
"LengthStalkedMaxCenter",
"LengthMin",
"LengthSwarmerMaxCenter",
"LengthSwarmerMin",
"LengthSwarmerMaxPole",
"WidthStalkedMaxPole",
"WidthStalkedMin",
"WidthStalkedMaxCenter",
"WidthMin",
"WidthSwarmerMaxCenter",
"WidthSwarmerMin",
"WidthSwarmerMaxPole",
"WidthStalkedMax",
"WidthSwarmerMax",
"WidthMax",
"LengthStalkedMax",
"LengthSwarmerMax",
"LengthMax",
"WidthsSmoothed",
)
for k in new_keys:
df[k] = None
# Evaluate each frame independently
for v in df.index:
# Assign default values
y = np.ndarray(0)
ast = np.nan
lstmaxp = np.nan
lstmin = np.nan
lstmaxc = np.nan
lst = np.nan
lswmaxc = np.nan
lswmin = np.nan
lswmaxp = np.nan
wstmaxp = np.nan
wstmin = np.nan
wstmaxc = np.nan
wmin = np.nan
wswmaxc = np.nan
wswmin = np.nan
wswmaxp = np.nan
wstmax = np.nan
wswmax = np.nan
wmax = np.nan
lstmax = np.nan
lswmax = np.nan
lmax = np.nan
try:
s = df.ix[v]
# Load variables from file
a = s["Area"]
l = s["Length"]
ps = (s["StalkedPole"], s["SwarmerPole"])
w = s["Widths"]
tcke = s["EdgeSpline"]
tckm = s["MidSpline"]
# Reconstruct cell contour and midline
n = len(w)
u = np.linspace(0.0, 1.0, n)
es = np.asarray(zip(*splev(u, tcke)))
ms = np.asarray(zip(*splev(u, tckm)))
dms = np.asarray(zip(*splev(u, tckm, der=1)))
# Smooth the width profile
y = gaussian_filter1d(w, sigma=sigma)
# Truncate the width profile near the poles
# y = y[m:-m]
# Find all relative minima and maxima
m = int(n * cutoff)
imin = np.hstack(argrelmin(y, order=m))
imax = np.hstack(argrelmax(y, order=m))
if np.any(imin):
# Find the smallest (primary) minimum
jmin = imin[np.argmin(y[imin])]
# Assign the minimum width
wmin = y[jmin]
# Find the normal line that split the cell at the minimum width
esmin = get_divn_plane(es, ms[jmin], dms[jmin])
# Split into stalked and swarmer parts
ast, _ = split_areas(u, tcke, esmin, es, ps, a)
lst = get_length(u[: jmin + 1], tckm)
# Sort the other minima to the stalked or swarmer part
istmin = imin[imin < jmin]
iswmin = imin[imin > jmin]
# Sort the maxima to the stalked or swarmer part
istmax = imax[imax < jmin]
iswmax = imax[imax > jmin]
if np.any(istmin):
# Find the smallest of the stalked minima
jstmin = istmin[np.argmin(y[istmin])]
# Assign the minimum stalked width
wstmin = y[jstmin]
# Find the length at the minimum stalked width
lstmin = get_length(u[: jstmin + 1], tckm)
# Sort the stalked maxima to the pole or center
istmaxp = istmax[istmax < jstmin]
istmaxc = istmax[istmax > jstmin]
# Find the largest of the stalked maxima
jstmaxp = istmaxp[np.argmax(y[istmaxp])]
jstmaxc = istmaxc[np.argmax(y[istmaxc])]
# Assign the stalked pole width and length maxima
wstmaxp = y[jstmaxp]
lstmaxp = get_length(u[: jstmaxp + 1], tckm)
# Assign the stalked center width and length maxima
wstmaxc = y[jstmaxc]
lstmaxc = get_length(u[: jstmaxc + 1], tckm)
else:
# Find the largest of the stalked maxima
jstmax = istmax[np.argmax(y[istmax])]
# Assign the stalked pole width and length maxima
wstmax = y[jstmax]
lstmax = get_length(u[: jstmax + 1], tckm)
if np.any(iswmin):
# Find the smallest of the swarmer minima
jswmin = iswmin[np.argmin(y[iswmin])]
# Assign the minimum swarmer width
wswmin = y[jswmin]
# Find the length at the minimum swarmer width
lswmin = get_length(u[: jswmin + 1], tckm)
# Sort the swarmer maxima to the pole or center
iswmaxp = iswmax[iswmax > jswmin]
iswmaxc = iswmax[iswmax < jswmin]
# Find the largest of the swarmer maxima
jswmaxp = iswmaxp[np.argmax(y[iswmaxp])]
jswmaxc = iswmaxc[np.argmax(y[iswmaxc])]
# Assign the swarmer pole width and length maxima
wswmaxp = y[jswmaxp]
lswmaxp = get_length(u[: jswmaxp + 1], tckm)
# Assign the swarmer center width and length maxima
wswmaxc = y[jswmaxc]
lswmaxc = get_length(u[: jswmaxc + 1], tckm)
else:
# Find the largest of the swarmer maxima
jswmax = iswmax[np.argmax(y[iswmax])]
# Assign the swarmer pole width and length maxima
wswmax = y[jswmax]
lswmax = get_length(u[: jswmax + 1], tckm)
else:
# Find the largest of the maxima
jmax = imax[np.argmax(y[imax])]
# Assign the width and length maxima
wmax = y[jmax]
lmax = get_length(u[: jmax + 1], tckm)
except Exception as e:
pass
s["WidthsSmoothed"] = y
s["AreaStalked"] = ast
s["LengthStalkedMaxPole"] = lstmaxp
s["LengthStalkedMin"] = lstmin
s["LengthStalkedMaxCenter"] = lstmaxc
s["LengthMin"] = lst
s["LengthSwarmerMaxCenter"] = lswmaxc
s["LengthSwarmerMin"] = lswmin
s["LengthSwarmerMaxPole"] = lswmaxp
s["WidthStalkedMaxPole"] = wstmaxp
s["WidthStalkedMin"] = wstmin
s["WidthStalkedMaxCenter"] = wstmaxc
s["WidthMin"] = wmin
s["WidthSwarmerMaxCenter"] = wswmaxc
s["WidthSwarmerMin"] = wswmin
s["WidthSwarmerMaxPole"] = wswmaxp
s["WidthStalkedMax"] = wstmax
s["WidthSwarmerMax"] = wswmax
s["WidthMax"] = wmax
s["LengthStalkedMax"] = lstmax
s["LengthSwarmerMax"] = lswmax
s["LengthMax"] = lmax
df.ix[v] = s
# Save new Series to file
for k in new_keys:
df[k].to_pickle(os.path.join(input_dir, k + ".pickle"))
