Ad-hoc QC tests for micro-CT systems have been established, using three commercial imaging phantoms (QRM GmbH, Möhrendorf, Germany) (Table 1).
Phantoms are acquired by micro-CT monthly25–28,31and segmentation maps with fixed regions of interest (ROIs) are applied to the scans for the assessment of the image quality parameters (Figure 1). These measurements are conducted in a grey levels scale, that corresponds to the original scale of CT scans after reconstruction.
The data reported in this study were collected on specific control charts and compared to baseline (BL) values, each defined by upper and lower tolerance limits, as suggested by literature25,31. BL values were initially established for our Quantum GX micro-CT (PerkinElmer, Inc. Waltham, MA), acquiring five consecutive scans of the QC phantoms31,32. For each monitored parameter, the BL and tolerance width were calculated, respectively:
$$BL= Average\left({\left({x}_{j}\right)}_{j=1,...,5}\right)$$
$$Tolerance width=2*SD\left({\left({x}_{j}\right)}_{j=1,\dots ,5}\right)$$
in which \(x\) represents an image quality parameter, \(j\) is the index of weekly acquisition and SD is the standard deviation of the baseline measures. All measures obtained by the QC tests were then compared to the corresponding tolerance range (Table 2):
$$Tolerance range=BL\pm Tolerance width$$
In this study, the dosimetry of the Quantum GX scanner33 and its impact on lung imaging were not objects of investigation, since they have been already evaluated in preclinical applications34–36, with findings suggesting that animal welfare was protected in this respect and that longitudinal CT data were not affected by multiple X-ray expositions.
1. Micro-CT phantoms and corresponding QC tests.
Water, Low Contrast and Bar Pattern phantoms (QRM GmbH, Möhrendorf, Germany) were used for quality control tests, as detailed below.
1.1 Water phantom
A cylindric device with a diameter of 32 mm and filled with milli-Q water, was employed to evaluate noise, to extract the absolute grey value for water and measure uniformity (Micro-CT water phantom, QRM GmbH, Möhrendorf, Germany).
Noise test. The noise value was extracted applying a circular ROI (ROINoise) to five spatially contiguous reconstructed slices, along the longitudinal z-axis (Figure 1A). The ROI area (80.4 mm2) was chosen to be 10% of the cylinder base area (16 mm x 16 mm x π = 804 mm2), according to IPEM and IEC indications37,38. The noise was defined as the SD of the water grey level (i.e. SDwater) within the ROI (i.e. meanwater ± SDwater). In order to obtain the monthly noise value, namely \({Noise}_{}\), we calculated the average of the five SDwater:
\(Noise=\sum _{i=1}^{5}\frac{\left({SD}_{i}^{water}\right)}{5}\) i = slice index
Water evaluation. Water grey level was monitored applying the same ROI employed for the noise test (Figure 1A). In this case, we evaluated the mean water grey level value (meanwater), averaging the values of the five contiguous cross sections that compose the ROIs:
\(Water=\sum _{i=1}^{5}\frac{\left({mean value}_{i}^{water}\right)}{5}\) i = slice index
Uniformity test. Five circular ROIs were positioned in the centre and in peripheral locations of the image and propagated for five contiguous slices along z-axis, as shown in Figure 1B. The image quality parameter for the uniformity test, i.e. \(Uniformity\), was calculated as the difference between the meanwater value of the central ROIa and the average of the four meanwater values of the peripheral ROIs (ROIb, ROIc, ROId, ROIe):
\(Uniformity={mean value}_{{ROI}_{a}}-\sum _{h=b}^{e}\frac{\left({mean value}_{{ROI}_{h}}\right)}{5}\) h = ROI index
Each \({mean value}_{ROI}\) was calculated by averaging the mean of water grey values obtained from the five contiguous slices, as in the ‘Water evaluation’ method.
1.2 Micro-CT Low Contrast phantom
The low contrast phantom was chosen to measure the contrast resolution of the scanner. It provides three approximate contrast levels of -9%, -6% and -3% compared to the background material, each level composed of three circular inserts with different diameters, i.e. 1, 2 and 3 mm, for a total of nine inserts (Micro-CT Low Contrast Phantom V2, QRM GmbH, Möhrendorf).
Low contrast test. The low contrast detectability (LCD) is defined as the minimum visible dimension for low contrast objects, provided by the diameter of the smallest circular insert. The segmentation map for the LCD test consists of nine circular ROIs with dimensions proportional to the three different sizes of the inserts (Figure 1D). After adjusting the position of the ROIs following the location of the inserts, the nine ROIs were propagated on 100 image slices to obtain a 3D evaluation. Two additional circular ROIs were also considered: one for resin background and one for air (outside the phantom). The image quality parameter for the LCD test, called \(Contrast\) and expressed in %, was calculated as follows:
$${Contrast}_{k}=\left[1-\frac{\left({mean value}_{k}-{mean value}_{air}\right)}{\left({mean value}_{background}-{mean value}_{air}\right)}\right]*100$$
where k represents the insert index and ranges from one to nine (k=1,…,9). Finally, for each contrast level (-9%, -6%, -3%), the \(Contrast\) parameter was evaluated as the average value of the three inserts with the same low contrast level but different diameter sizes:
$${Contrast}^{c}=(\frac{{Contrast}_{1mm}^{c}+{Contrast}_{2mm}^{c}+{Contrast}_{3mm}^{c})}{3}$$
in which the index c can be -9%, -6% or -3%. In this way, each month we obtained three values of \(Contrast\), one for each nominal level.
1.3 Micro-CT Bar Pattern phantom
The spatial resolution test, as well as the evaluation of the absolute grey value for air, were carried out with the bar pattern phantom filled with air, and with a diameter of 20 mm. Two chips placed inside the phantom contain bar patterns with different widths and points with different diameters allowed evaluation of the spatial resolution in the centre as well as in the periphery of the image in a single measurement (Micro-CT Bar Pattern Phantom, QRM GmbH, Möhrendorf).
Spatial resolution test. The in-plane spatial resolution for micro-CT images was tested. Based on visual inspection of the line pairs in the phantom scan (Figure 1E), the spatial resolution was in a range between 10 and 3.3 line pairs per mm (lp/mm), corresponding to 50 to 150 µm line width (Figure 1F), which represents the range in which the bars can be resolved26. The Modulation Transfer Function (MTF) of a system is used to describe the resolution performance of the scanner, providing a quantitative measure of the relationship between the original object and its radiological reconstruction39. MTF was calculated monthly for the three different bar patterns (3.3 lp/mm, 5 lp/mm and 10 lp/mm) using the ‘line profile’ tool of Analyze software (Analyze 12.0; Copyright 1986–2017, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN), https://www.analyzedirect.com. The value averaged over three measures for ‘background’ and ‘bars’ was recorded and MTF calculated as follow:
$${MTF}_{}^{c}=\left[\frac{\left({mean value}_{background}-{mean value}_{bars}\right)}{\left({mean value}_{background}+{mean value}_{bars}\right)}\right]*100$$
where the index c can be 3.3 lp/mm, 5 lp/mm or 10 lp/mm. Since axial resolution is the same as transversal resolution, only the latter data are reported.
Air evaluation. The monitoring of grey levels of air was performed using a circular ROI positioned inside the bar pattern phantom (in air), with an area of 31.4 mm2, thus covering 10% of the phantom section area (10 mm x 10 mm x π = 314 mm2). The ROI was propagated for five contiguous cross sections and the mean grey level of air (meanair) was calculated as the average of the five values extracted, as following:
\(Air=\sum _{i=1}^{5}\frac{\left({mean value}_{i}^{air}\right)}{5}\) i = slice index
2. Experimental control animals
A cohort of 22 control mice were used for a retrospective evaluation of the impact of routine QC tests on lung scans post-processing.
Female inbred C57Bl/6 (7- to 8- weeks old) mice were purchased from Envigo, Italy (San Pietro al Natisone, Udine, Italy). Prior to use, animals were acclimatized for at least 5 days to the local vivarium conditions (room temperature: 20–24°C; relative humidity: 40–70%; 12-h light–dark cycle), having free access to standard rodent chow and softened tap water.
All animal experiments described herein were authorized by the official competent authority and approved by the intramural animal-welfare body (AWB) of Chiesi Farmaceutici and authorized by the Italian Ministry of Health (protocol number: 809/2020-PR). All procedures were conducted in an AAALAC (Association for Assessment and Accreditation for Laboratory Animal Care) certified facility in compliance with the European Directive 2010/63 UE, Italian D.Lgs 26/2014, the revised “Guide for the Care and Use of Laboratory Animals”40 and with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines41.
Animals were lightly anesthetized with 2.5% isoflurane delivered in a box and vehicle (50 µl saline [0.9%]) was administered via oropharyngeal aspiration (OA) using a micropipette42. Twenty-one days after saline administration, mice were anesthetized with 2% isoflurane and underwent micro-CT imaging10,14.
3. Micro-CT settings, images acquisition and processing
The Quantum GX Micro-CT (PerkinElmer, Inc. Waltham, MA) was used in this study. This scanner has a microfocus X-ray source with a Tungsten anode. A fixed filter of 0.5 mm aluminium (Al) and 0.06 mm copper (Cu) is placed in front of the exit port to remove low energy X-rays. QC phantoms were acquired with the following parameters: X-ray tube voltage 90 KV, X-ray tube current 88 µA, total scan time of 4 min. A ring reduction correction was applied to the sinograms and the entire set of projection radiographs was input into a GPU-based filtered back-projection algorithm with a Ram-Lak filter. The acquisition protocol allows acquisition of projections over a total angle of 360° resulting in 3D datasets with 50 µm isotropic reconstructed voxel size10,43. All the images were imported as 3D .vox files and analysed using Analyze software (Analyze 12.0; Copyright 1986–2017, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN) https://www.analyzedirect.com.
All protocols for mouse lung acquisition and image post-processing were largely detailed by Mecozzi et al.10 applied to a bleomycin-induced fibrosis murine model. Lung density histograms were extracted from pulmonary scans and the Area Under Curve (AUC), the skewness and kurtosis calculated.
4. Statistical Analysis
Statistical analyses were performed using GraphPad Prism version 9.1.2 for Windows (GraphPad Software, La Jolla, CA, USA), https ://www.graphpad.com. All data were presented as mean ± SD. The normality test was performed for the AUC, the skewness and kurtosis of the average lung histograms. A one-tailed unpaired t-test was performed to compare AUC parameters. Finally, an unpaired t-test and Mann-Whitney test were performed to compare the kurtosis and the skewness, respectively. The alpha level of all tests was set at 0.05.