Radiation dose optimized lateral expansion of the field of view in synchrotron radiation X-ray tomographic microscopy

Abstract

Volumetric data at micrometer level resolution can be acquired within a few minutes using synchrotron-radiation-based tomographic microscopy. The field of view along the rotation axis of the sample can easily be increased by stacking several tomograms, allowing the investigation of long and thin objects at high resolution. On the contrary, an extension of the field of view in the perpendicular direction is non-trivial. This paper presents an acquisition protocol which increases the field of view of the tomographic dataset perpendicular to its rotation axis. The acquisition protocol can be tuned as a function of the reconstruction quality and scanning time. Since the scanning time is proportional to the radiation dose imparted to the sample, this method can be used to increase the field of view of tomographic microscopy instruments while optimizing the radiation dose for radiation-sensitive samples and keeping the quality of the tomographic dataset on the required level. This approach, dubbed wide-field synchrotron radiation tomographic microscopy, can increase the lateral field of view up to five times. The method has been successfully applied for the three-dimensional imaging of entire rat lung acini with a diameter of 4.1 mm at a voxel size of 1.48 microm.

Figure 1

Covering the field of view of differently sized samples with one 180° scan (a), one 360° scan (b) or, in the case of the so-called wide-field scanning, with multiple subscans (three subscans, c). The filled segments mark the region of the sample that is covered while scanning the respective positions (position 1: magenta/checkerboard; position 2: yellow; position 3: cyan/striped).

Figure 2

Wide-field scan set-up with three 180° scans; one central (yellow) and two lateral scans (magenta and cyan or top and bottom, respectively). In this drawing, four projections for the central and eight projections for each of the lateral scans have been recorded. The colors of the three positions correspond to the colors shown in Fig. 1(c) ▶. (a) Scanned projections; (b) scanned projections and additional interpolated projections (dotted) required to merge all projections.

Figure 3

Set-up for different fields of view. (a) Desired field of view of 3072 pixel diameter. (b) Wide-field scanning protocol for covering the desired field of view of panel (a) with merged projections from one central and two half ring scans (r ₁ and r ₂). (c) Desired field of view of 5120 pixel diameter. (d) Wide-field scanning protocol for covering the desired field of view of panel (c) with merged projections from one central and four half ring scans (r ₁–r ₄). (e) Desired field of view of 7168 pixel diameter. (f) Wide-field scanning protocol for covering the desired field of view of panel (e) with merged projections from one central and six half ring scans (r ₁–r ₆).

Figure 4

Workflow of a wide-field scan. The images show a rat lung sample from a Sprague-Dawley rat, obtained 21 days after birth, scanned with the acquisition protocol B (Table 1 ▶). (a) Three corrected and independently acquired projections from subscans s ₁–s ₃ are shown. Each one is 1024 × 1024 pixels large and covers a field of view of 1.52 mm. Subscans s ₁ and s ₂ overlap by 141 pixels (red and green overlay), subscans s ₂ and s ₃ overlap by 138 pixels (blue and yellow overlay). (b) Merged projection obtained from the three subscans shown in subfigure (a). Each merged projection has a size of 2792 × 1024 pixels. Owing to the overlap required to merge the projections, the width of the merged projections is slightly smaller than three times the width of the subscans. (c) Cropped slice of the reconstructed tomographic dataset. The dashed red circles mark the start and end of the overlap region.

Figure 5

Three-dimensional visualization of the distal-medial tip of the right lower rat lung lobe. The gray structure in the background shows a semi-transparent view of the tomographic dataset with segmented airways. The foreground shows isosurfaces of terminal airways. The wireframe cube has a side length of 1024 pixels and encloses the field of view of one conventional scan. (a) Conventional scan; the extracted airway segments (magenta and yellow or left and right, respectively) are only partially contained inside the total sample volume. Airway segments not contained in the dataset but present in the sample are shown semi-transparent. This conventional scan corresponds to a reconstruction of the central of the three wide-field scan subscans. (b) Wide-field scan with increased field of view; the magenta (center) and yellow segment (right) show entire acini inside the dataset; the cyan segment (left) contains a partially cut acinus. All airway segments inside the sample are contained in the tomographic dataset.

Figure 6

Plot of normalized difference value (, blue diamonds) for the 19 scanned protocols overlaid over quality-plot (red dots) obtained from the simulation (described in §2.4). The normalized error has been calculated using the difference image of each protocol i with protocol B. The error bars for each protocol show the standard deviation of the error calculated for 205 of the 1024 slices. Note that the scale of the error was normalized to 20–100%, so that both the quality from the simulation and the error are directly comparable. The abscissa shows the scanning time in percentage of time used for the gold standard scan. Protocol T on the far left corresponds to the fastest scanning time, protocol B on the far right to the slowest. The protocols in between are shown from T–B for increasing percentage of the scanning time.

Figure 7

Comparison of three-dimensional visualizations. (a), (b), (c) Three independent airway segments (cyan, magenta, yellow) of tomographic datasets obtained with protocol B, L and T, extracted using a region-growing algorithm. A cubic ROI (blue) with a side length of 256 pixels (corresponding to 379 µm) is marked inside the leftmost segment for all protocols. (d), (e), (f) Detailed view of isosurfaces of the lung tissue inside the blue ROIs for protocol B, L and T, respectively. Note the increasing surface roughness in the alveolar surfaces for subfigures (e) and (f).

Figure 8

Overview of the location of the four ROIs where the histogram of the euclidean distance transformation distribution has been calculated. Gray: semi-transparent volume rendering of the lung tissue sample. Red: four ROIs, extracted to calculate the distance transformation. The labels of the ROIs conform to the legends in Fig. 9 ▶.

Figure 9

Histogram plots for each of the four ROIs, each showing the histogram of the distance transformation for the protocols B, L and T.