Example

In [1]:

%matplotlib widget
import numpy as np
import matplotlib.pyplot as plt
import fibretracker as ft

%matplotlib widget import numpy as np import matplotlib.pyplot as plt import fibretracker as ft

In [2]:

data_path = '/Users/pooja/Documents/PhD_work/ICPR_docs/data/Mock_UD_GFRP_data/UD-01_FoV_2_B2_recon.txm'
data_path1 = 'data/Mock_UD_GFRP_data/Mock-01_FoV_2_B2_recon.txm'
data_path2 = 'data/Mock_UD_GFRP_data/GFRP'
data_path3 = 'data/Mock_UD_GFRP_data/XCT_LR'

V_UD = ft.io.load(data_path)
# V_Mock = ft.io.load(data_path1)
# V_GFRP = ft.io.load(data_path2, contains='.tif')
# V_XCTLR = ft.io.load(data_path3, contains='.tif')

V_norm = ft.io.normalize(V_UD).astype(float)

data_path = '/Users/pooja/Documents/PhD_work/ICPR_docs/data/Mock_UD_GFRP_data/UD-01_FoV_2_B2_recon.txm' data_path1 = 'data/Mock_UD_GFRP_data/Mock-01_FoV_2_B2_recon.txm' data_path2 = 'data/Mock_UD_GFRP_data/GFRP' data_path3 = 'data/Mock_UD_GFRP_data/XCT_LR' V_UD = ft.io.load(data_path) # V_Mock = ft.io.load(data_path1) # V_GFRP = ft.io.load(data_path2, contains='.tif') # V_XCTLR = ft.io.load(data_path3, contains='.tif') V_norm = ft.io.normalize(V_UD).astype(float)

astropy module not found

In [3]:

ft.viz.slicer(V_norm, img_height=7, img_width=7, cmap='gray')

ft.viz.slicer(V_norm, img_height=7, img_width=7, cmap='gray')

Out[3]:

interactive(children=(IntSlider(value=499, description='Slice', max=998, style=SliderStyle(description_width='…

Figure

In [4]:

V_tmp = V_norm[100:350, V_norm.shape[1]//2 - 250:V_norm.shape[1]//2 + 250, V_norm.shape[2]//2 - 250:V_norm.shape[2]//2 + 250]

V_tmp = V_norm[100:350, V_norm.shape[1]//2 - 250:V_norm.shape[1]//2 + 250, V_norm.shape[2]//2 - 250:V_norm.shape[2]//2 + 250]

In [5]:

ft.viz.slicer(V_tmp, img_height=7, img_width=7, cmap='gray')

ft.viz.slicer(V_tmp, img_height=7, img_width=7, cmap='gray')

Out[5]:

interactive(children=(IntSlider(value=125, description='Slice', max=249, style=SliderStyle(description_width='…

Figure

In [6]:

ft.viz.orthogonal(V_tmp, img_height=7, img_width=7, cmap='gray')

ft.viz.orthogonal(V_tmp, img_height=7, img_width=7, cmap='gray')

Out[6]:

interactive(children=(HBox(children=(IntSlider(value=125, description='Z', max=249), IntSlider(value=250, desc…

Figure

In [7]:

n_bins = int(255*(V_tmp.max() - V_tmp.min()) + 1)
hist, bins = np.histogram(V_tmp, bins=n_bins)
bin_width = bins[1] - bins[0]
fig, ax = plt.subplots(1, figsize=(5, 5))
ax.bar(bins[:-1], hist, width=bin_width)
ax.set_xlabel('Intensity')
ax.set_ylabel('Frequency')
ax.set_title('Histogram of UD section volume')
plt.show()

n_bins = int(255*(V_tmp.max() - V_tmp.min()) + 1) hist, bins = np.histogram(V_tmp, bins=n_bins) bin_width = bins[1] - bins[0] fig, ax = plt.subplots(1, figsize=(5, 5)) ax.bar(bins[:-1], hist, width=bin_width) ax.set_xlabel('Intensity') ax.set_ylabel('Frequency') ax.set_title('Histogram of UD section volume') plt.show()

Figure

parameters of std (gaussian filter size - more noise higher the value), min_distance (based on fibre size in scan - higher pixel diameter; higher the value) and threshold_abs [value seperating matrix and fibre from histrogram] need to be tuned so that most of the fibre are accurately detected.

In [9]:

coords = ft.models.get_fibre_coords(V_tmp, std=2.5, min_distance=3, threshold_abs=0.6)

coords = ft.models.get_fibre_coords(V_tmp, std=2.5, min_distance=3, threshold_abs=0.6)

Slide few slices using slider widget to check most of the fibre are detected across slices

In [10]:

ft.viz.slicer(V_tmp, detect_coords=coords, mark_size=3, img_height=7, img_width=7, cmap='gray')

ft.viz.slicer(V_tmp, detect_coords=coords, mark_size=3, img_height=7, img_width=7, cmap='gray')

Out[10]:

interactive(children=(IntSlider(value=125, description='Slice', max=249, style=SliderStyle(description_width='…

Figure

In [11]:

# fibre can be tracked either directly from the volume or from the detected coordinates, or both[to avoid coordinate detection again]

tracks = ft.models.track_fibres(coords=coords)
tracks_gauss = ft.models.track_fibres(V_tmp, coords=coords, smoothtrack_gaussian=True, sigma=2)
tracks_ws = ft.models.track_fibres(V_tmp, coords=coords, smoothtrack_watershed=True, threshold=0.6)

# tracks = ft.models.track_fibres(V_tmp, smoothtrack_watershed=True, threshold=0.5)
# tracks = ft.models.track_fibres(V_tmp, smoothtrack_gaussian=True)
# tracks = ft.models.track_fibres(coords=coords)
# tracks = ft.models.track_fibres(coords=coords, smoothtrack_gaussian=True)

# fibre can be tracked either directly from the volume or from the detected coordinates, or both[to avoid coordinate detection again] tracks = ft.models.track_fibres(coords=coords) tracks_gauss = ft.models.track_fibres(V_tmp, coords=coords, smoothtrack_gaussian=True, sigma=2) tracks_ws = ft.models.track_fibres(V_tmp, coords=coords, smoothtrack_watershed=True, threshold=0.6) # tracks = ft.models.track_fibres(V_tmp, smoothtrack_watershed=True, threshold=0.5) # tracks = ft.models.track_fibres(V_tmp, smoothtrack_gaussian=True) # tracks = ft.models.track_fibres(coords=coords) # tracks = ft.models.track_fibres(coords=coords, smoothtrack_gaussian=True)

Smoothing tracks using Gaussian       
Smoothing tracks using watershed...   
Watershed volume created.
Smoothing tracks - iteration: 250/250

In [12]:

ft.viz.plot_tracks(tracks)

ft.viz.plot_tracks(tracks)

Figure

Out[12]:

(<Figure size 1000x1000 with 2 Axes>, <Axes3D: >)

In [13]:

ft.viz.plot_tracks(tracks_gauss)

ft.viz.plot_tracks(tracks_gauss)

Figure

Out[13]:

(<Figure size 1000x1000 with 2 Axes>, <Axes3D: >)

In [14]:

ft.viz.plot_tracks(tracks_ws)

ft.viz.plot_tracks(tracks_ws)

Figure

Out[14]:

(<Figure size 1000x1000 with 2 Axes>, <Axes3D: >)