Critical point drying of brain tissue for X-ray phase-contrast imaging.

1. # %% [markdown]
2. # # ID16A Line-Profile Processing (Example)
3. #
4. # This notebook demonstrates the workflow for a **single** ID16A dataset (one CSV):
5. # 1. Plot the **raw** line profiles.
6. # 2. Compute and apply a **baseline correction** using specified baseline windows.
7. # 3. Plot the **averaged corrected** line profile with ±1 standard deviation.
8. # 4. Export the averaged profile + std to CSV.
9. #
10. # - Column 0: distance in **pixels** → converted to µm using `0.100 µm/pixel`.
11. # - Intensity columns: every other column starting from column 1.
13. # %%
14. import os
15. import numpy as np
16. import pandas as pd
17. import matplotlib.pyplot as plt
18. from matplotlib.ticker import ScalarFormatter
20. # ==== Parameters ====
21. FILE_PATH = "data/ID16A/ID16A_Epon_lineprofiles_SK186_p2_n100.csv" # <-- set this to one CSV file
23. PIXELS_TO_UM = 0.100 # µm per pixel for ID16A (100 nm)
24. SAVE_FIGS = True
25. FIG_DPI = 300
27. EPS = 1e-12 # small number to avoid division-by-zero if ever needed
29. # %%
30. # defaults
31. plt.rcParams.update({
32. "font.size": 14,
33. "xtick.direction": "in",
34. "ytick.direction": "in",
35. "xtick.top": True,
36. "ytick.right": True,
37. "axes.linewidth": 1.5,
38. "figure.dpi": 150
39. })
41. # %%
42. def safe_read_csv(file_path):
43. """Safely read a CSV file with basic error handling."""
44. if not os.path.exists(file_path):
45. raise FileNotFoundError(f"File not found: {file_path}")
46. return pd.read_csv(file_path, header=None, low_memory=False)
49. def build_baseline_mask(dist_um):
50. """Return a boolean mask for baseline regions with fixed windows in µm."""
51. baseline_windows_um = [(0, 1.8), (6.2, 8)]
52. mask = np.zeros_like(dist_um, dtype=bool)
53. for lo, hi in baseline_windows_um:
54. mask |= (dist_um >= lo) & (dist_um <= hi)
55. return mask
58. def extract_profiles(df, pixels_to_um=0.100):
59. """
60. From a raw dataframe:
61. - Convert the first column (pixels) -> distance in µm
62. - Take every other column starting at col 1 as profiles
63. - Multiply profiles by conversion factor to match P14 units (rad/voxel)
64. Returns: dist_um (N,), profiles (N, M) in P14 units
65. """
66. dist_um = df.iloc[:, 0].to_numpy(dtype=float) * pixels_to_um
67. profiles = df.iloc[:, 1::2].to_numpy(dtype=float)
69. # --- Conversion factor (per cm → rad/voxel as in P14 units) [see text below]---
70. conversion_factor = 0.512 * 0.325 * 1e-4
71. profiles = profiles * conversion_factor
73. return dist_um, profiles
76. def baseline_correct_profiles(dist_um, profiles):
77. """
78. Subtract per-profile baseline mean using fixed windows:
79. - 0 → 1.8 µm
80. - 6.2 → 8 µm
81. Returns: corrected_profiles (N, M), baseline_means (M,)
82. """
83. base_mask = build_baseline_mask(dist_um)
84. corrected = np.empty_like(profiles)
85. baseline_means = np.empty(profiles.shape[1])
86. for i in range(profiles.shape[1]):
87. base_mean = np.mean(profiles[base_mask, i])
88. baseline_means[i] = base_mean
89. corrected[:, i] = profiles[:, i] - base_mean
90. return corrected, baseline_means
93. def compute_avg_and_std(corrected_profiles):
94. """Compute mean and std across profiles (axis=1)."""
95. mean_profile = np.mean(corrected_profiles, axis=1)
96. std_profile = np.std(corrected_profiles, axis=1, ddof=0)
97. return mean_profile, std_profile
99. # %% [markdown]
100. # ### Unit Conversion (ID16A → P14)
101. #
102. # At ID16A, the holotomographic reconstruction provides values as:
103. #
104. # $$
105. # \omega = - \frac{2\pi}{\lambda} \, \delta
106. # $$
107. #
108. # in units of **cm$^{-1}$**.
109. #
110. # To compare directly with P14 (which reports in **rad/voxel**), each line profile
111. # is multiplied by the conversion factor:
112. #
113. # $$
114. # \text{factor} = 0.512 \times 0.325 \times 10^{-4}
115. # $$
116. #
117. # where 0.512 arrises from the the ratio of the X-ray wavelength (energies) at P14 to ID16A, and 0.325 $\times 10^{-4}$ is the pixel size at P14 in per cm units.
118. #
119. # This converts ID16A values into the same effective units as P14.
121. # %%
122. df = safe_read_csv(FILE_PATH)
123. dist_um, profiles = extract_profiles(df, pixels_to_um=PIXELS_TO_UM)
125. print(f"Loaded: {FILE_PATH}")
126. print(f"Distance array length: {dist_um.size}, profiles shape: {profiles.shape}")
128. # %%
129. fig, ax = plt.subplots(figsize=(7, 5))
130. for i in range(profiles.shape[1]):
131. ax.plot(dist_um, profiles[:, i], alpha=0.3, linewidth=1)
133. ax.axhline(0, linestyle='--', linewidth=1)
134. ax.set_xlabel("Distance (µm)")
135. ax.set_ylabel("relative phase-shift (rad/voxel) [raw signal]")
136. ax.set_xlim(0, 8)
137. ax.set_ylim(-1e-2, 1e-3)
139. # Force scientific notation on y-axis if needed
140. fmt = ScalarFormatter(useMathText=True)
141. fmt.set_scientific(True)
142. fmt.set_powerlimits((0, 0))
143. ax.yaxis.set_major_formatter(fmt)
145. ax.tick_params(axis='both', length=6, width=1.5)
146. for spine in ax.spines.values():
147. spine.set_linewidth(1.5)
149. plt.tight_layout()
150. if SAVE_FIGS:
151. plt.savefig("ID16A_raw_profiles.svg", format='svg', dpi=FIG_DPI, bbox_inches='tight')
152. plt.show()
154. # %%
155. corrected_profiles, baseline_means = baseline_correct_profiles(dist_um, profiles)
156. print(f"Computed baseline means for {profiles.shape[1]} profiles. "
157. f"Example (first 5): {baseline_means[:5]}")
159. # %%
160. fig, ax = plt.subplots(figsize=(7, 5))
161. for i in range(corrected_profiles.shape[1]):
162. ax.plot(dist_um, corrected_profiles[:, i], alpha=0.3, linewidth=1)
164. ax.axhline(0, linestyle='--', linewidth=1)
165. ax.set_xlabel("Distance (µm)")
166. ax.set_ylabel("Corrected phase-shift (rad/voxel)")
167. ax.set_xlim(0, 8)
168. ax.set_ylim(-1e-2, 1e-3)
170. fmt = ScalarFormatter(useMathText=True)
171. fmt.set_scientific(True)
172. fmt.set_powerlimits((0, 0))
173. ax.yaxis.set_major_formatter(fmt)
175. ax.tick_params(axis='both', length=6, width=1.5)
176. for spine in ax.spines.values():
177. spine.set_linewidth(1.5)
179. plt.tight_layout()
180. if SAVE_FIGS:
181. plt.savefig("ID16A_corrected_profiles.svg", format='svg', dpi=FIG_DPI, bbox_inches='tight')
182. plt.show()
184. # %%
185. mean_profile, std_profile = compute_avg_and_std(corrected_profiles)
187. fig, ax = plt.subplots(figsize=(7, 5))
188. ax.plot(dist_um, mean_profile, linewidth=2, label='Mean profile')
189. ax.fill_between(dist_um, mean_profile - std_profile, mean_profile + std_profile,
190. alpha=0.15, label='±1σ')
192. ax.axhline(0, linestyle='--', linewidth=1)
193. ax.set_xlabel("Distance (µm)")
194. ax.set_ylabel("Corrected phase-shift (rad/voxel)")
195. ax.set_xlim(0, 8)
196. ax.set_ylim(-1e-2, 1e-3)
198. fmt = ScalarFormatter(useMathText=True)
199. fmt.set_scientific(True)
200. fmt.set_powerlimits((0, 0))
201. ax.yaxis.set_major_formatter(fmt)
203. ax.legend(frameon=False)
204. ax.tick_params(axis='both', length=6, width=1.5)
205. for spine in ax.spines.values():
206. spine.set_linewidth(1.5)
208. plt.tight_layout()
209. if SAVE_FIGS:
210. plt.savefig("ID16A_corrected_mean_with_std.svg", format='svg', dpi=FIG_DPI, bbox_inches='tight')
211. plt.show()
213. # %%
214. export_df = pd.DataFrame({
215. 'Distance (µm)': dist_um,
216. 'Mean Corrected Profile': mean_profile,
217. 'Standard Deviation': std_profile
218. })
219. out_csv = "ID16A_averaged_corrected_profile.csv"
220. export_df.to_csv(out_csv, index=False)
221. print(f"Saved: {out_csv}")

ID16A_Analysing_LineProfile_CPD-Paper.ipynb at commit 7087f77, under MIT · at the source