Abstract
Numerous studies have highlighted the role of X-ray computed tomography (X-ray CT) in understanding root architecture. Nevertheless, setting definitive scanning parameters for diverse soils in varied container sizes remains challenging. This study investigates the influence of X-ray CT system voltage on the penetration capability in diverse soils and container sizes, focusing on two key parameters: (1) gray values, which indicate X-ray attenuation and contribute to image contrast, and (2) signal-to-noise ratio, a measure of image clarity. Five soil samples were collected from various depths within a soil profile to encompass bulk density values ranging from 1.34 to 1.84 g·cm−3 to conduct the experiment. Containers with dimensions of 6 × 6 × 6 cm³, 8 × 8 × 6 cm³, 10 × 10 × 6 cm³, 12 × 12 × 6 cm³, 14 × 14 × 6 cm³, and 16 × 16 × 6 cm³ were used. Voltage levels spanning 75 to 225 kV, in 25-kV increments, were applied to each sample. The observed gray values of the X-ray images were fitted using a logistic model of three parameters. Results showed that increasing voltage leads to enhanced penetration up to a plateau point, irrespective of soil density or container size. This plateau could potentially yield higher quality scans, given that lower voltages result in subdued gray values and reduced image contrast. Notably, it was observed that soil properties, including mineral composition, directly affect image gray values. This study established optimal voltage settings for specific soil types at fixed densities, offering valuable insights for researchers investigating soil–root interactions. Although the current findings are based on five soils, a more extensive sampling encompassing diverse soil textures and densities is necessary for a comprehensive understanding of X-ray penetration behavior across various soil types.
Root architecture plays a pivotal role in enhancing crop yields, particularly amid the challenges posed by global population growth and the imperative for food security (Gregory and George 2011; Lynch 2022; Lynch et al. 2021). However, traditional breeding programs have predominately focused on optimizing aboveground plant growth, overlooking the intricate dynamics of below-ground plant systems, specifically root structures. Consequently, there is a pressing need for innovative approaches to analyze belowground crop growth comprehensively, particularly within field conditions (Araus et al. 2022). X-ray computed tomography (CT) has emerged as a transformative tool for phenotyping crop root systems, offering nonHodestructive, high-resolution, three-dimensional (3D) spatial, and temporal analyzes of roots (Hou et al. 2022; Li et al. 2022; Mooney et al. 2012). This technology empowers researchers to gain insights into root architecture, nutrient uptake, and water absorption dynamics, thereby facilitating the development of crops and crop systems with enhanced productivity and resilience.
The use of X-ray computed tomography (CT) as a phenotyping tool for crop root systems has witnessed remarkable growth in recent years (Hou et al. 2022). The nondestructive nature of X-ray CT enables researchers to conduct comprehensive 3D spatial and temporal analyzes of roots, offering higher resolution and time efficiency compared with alternative nondestructive techniques, such as magnetic resonance imaging (MRI) (Hou et al. 2022; Li et al. 2022; Mooney et al. 2012). X-ray CT has been extensively used in soil science for qualitative and quantitative analysis of soil properties, yet its application in horticultural container-based research has been sporadic and limited (Brown et al. 1987; Yafuso et al. 2019). As an investigative tool, X-ray CT has been successfully demonstrated in studies involving organic substrate-filled containers with small sample volumes, such as those commonly employed in nursery and floriculture production systems (Bartley 2022; Bartley et al. 2019; Yafuso et al. 2019). These studies demonstrated the potential of X-ray CT to observe dynamic processes within container substrates, shedding light on water distribution patterns, substrate characteristics, and plant–root interactions. Soilless container-based cultivation offers distinct advantages for the use of X-ray CT, such as low bulk densities and manageable sample sizes (Bartley et al. 2019). However, many horticultural crop investigations involve mineral soil-based production systems. Despite significant contributions from the scientific community, further research in this domain has been hindered by technical challenges and methodological limitations.
Thorough and detailed reviews of X-ray CT have been provided by Pires et al. (2010), Stock (2008), and Wildenschild et al. (2002). A CT scan involves passing monochromatic X-ray beams with incident intensity I0 through a sample of thickness τ, resulting in attenuation and a reduction in X-ray beam intensity, allowing for the detection of the material under consideration (Hou et al. 2022; Pires et al. 2010). A sensor detects and records the attenuated X-ray beam as a range of gray values, producing a series of 2D projections, which are subsequently reconstructed into a 3D dataset for further analysis. It is commonly observed that increasing the energy of an X-ray system, typically achieved by elevating the system voltage, enhances the penetration of the photons and, ultimately, signal-to-noise ratio (SNR), serving as an indirect measure of scan quality (Krbcova and Kukal 2017; Shikhaliev 2010). Therefore, researchers interested in using CT may tend to use a higher voltage of the system to obtain better scans. However, challenges such as beam hardening, causing excessive absorption of low-energy photons and resulting in streaks and cupping effects, can arise, particularly at higher voltages (Sprawls 1995). On the other hand, the “penetration plateau” or “kV plateau” refers to a point at which further increments in X-ray energy do not significantly improve penetration because sufficient energy is already available to penetrate most materials, leading to the saturation of X-ray penetration capability (Sprawls 1995). Identifying the penetration plateau point is crucial for obtaining the maximum gray values that material would need to achieve optimal scan quality for a given sample.
The quality of an X-ray CT scan is affected by several physical factors, including the size of the sample under analysis, the voltage of the X-ray beam, the density of the sample through which the X-ray passes, and the chemical composition of the material (Baveye et al. 2002; Khosravani and Reinicke 2020; Nazarian et al. 2008; Tollner and Murphy 1991). For instance, Gao et al. (2019) analyzed a root sample using 140 kV contained in a cylinder of 7 cm diameter and growing in a silt loam soil compacted to a bulk density of 1.27 g·cm−3, whereas Flavel et al. (2017) used a reduced voltage of 130 kV, owing to a smaller sample of 5.5 cm diameter and fine textured soil packed at a reduced density of 1.1 g·cm−3. Similarly, Kirk et al. (2019) analyzed a silt loam texture soil packed at 0.81 g·cm−3 in an 8 cm diameter container at 120 kV. Therefore, researchers have used a range of voltages for different soils compacted at different bulk densities, while no general guidelines are available to obtain a high-quality scan. The use of inaccurate voltage of the CT system would limit the amount of detail that can be extracted from the available scan.
Establishing definitive scanning parameters for soils packed in containers of different sizes will help efficiently obtain the best available scan for a sample and provide more insights from the scan. This study aimed to address this knowledge gap by investigating the optimization of X-ray CT parameters, with a specific focus on system voltage, for effective imaging of mineral soil-based systems in containers. By systematically varying system voltage and analyzing its impact on X-ray penetration and image quality, this investigation seeks to establish optimal scanning parameters tailored to different container sizes and soil types.
Materials and Methods
Soil and containers.
Five soil samples (Samples A–E) were collected from different depths showing distinct horizons within the Cowart series (fine-loamy, kaolinitic, thermic Typic Kanhapludults) soil profile located near the Soil Survey Office, Tuskegee, AL, USA (32°26′5.028″N, 85°43′50.196″W) to obtain a range of bulk density values. These samples were sieved through a 2-mm sieve to ensure uniformity and sent to Waters Agricultural Laboratories Inc. (Camila, GA, USA) for texture analysis. To facilitate the experimentation, six plexiglass rectangular prism containers with a thickness of ∼3.5 mm were used. The inner dimensions of these containers varied, with sizes of 6 × 6 × 6 cm³, 8 × 8 × 6 cm³, 10 × 10 × 6 cm³, 12 × 12 × 6 cm³, 14 × 14 × 6 cm³, and 16 × 16 × 6 cm³. The bulk density for each sample was determined by filling each container with dry soil, ensuring a uniformly packed sample, free of cracks.
X-ray CT settings.
The experiment was conducted using a Pinnacle X-ray Solutions (PXS) Benchmark 225/500 X-ray CT system with a final voxel size of 88 μm. The source-to-object distance (SOD) was maintained at 265 mm, wheras the source-to-detector distance (SDD) was maintained at 600 mm, resulting in a magnification factor of 2.26×. The X-ray tube was configured in 2 × 2 binning mode, resulting in a pixel size of 200 μm, and before each scan, the detector was calibrated. A range of voltages spanning from 75 to 225 kV with increments of 25 kV was employed for each sample. The current was adjusted individually for each voltage setting to attain a similar detector saturation level across all measurements. Furthermore, a copper filter of 1 mm thickness was used to prevent beam hardening by reducing low-energy X-rays passing through the sample and improving the image contrast following the approach suggested by Rana et al. (2015) and Shikhaliev (2005).
Data collection and analysis.
This experiment conducted X-ray CT scans by placing plexiglass containers filled with bulk soil samples on the CT stage and exposing them to X-rays. As the X-rays passed through the soil, a flat panel detector captured the transmitted rays, converting them into electrical signals. These signals were then processed to generate digital images, where each pixel was assigned a gray value reflecting the X-ray attenuation at that point in the soil. To extract these gray values, Pxsviewer software was employed. Essentially, these gray values represent the brightness of each point in the CT images, providing a measure of X-ray energy absorption by the soil. Higher gray values generally indicate lower X-ray absorption and potentially better image contrast. From each sample, four 1-cm2 regions surrounding the center were analyzed, and the average value was used for subsequent analysis. To mitigate any influence of a specific side, gray values were collected from two sides of the container by rotating the sample 90 degrees. Although this approach was not designed to generate 3-D reconstructions, it provided a more comprehensive assessment of X-ray attenuation through soils of varying bulk densities. This methodology establishes a foundation for future studies involving root imaging by enabling better differentiation between roots and the surrounding soil matrix.
Where the parameters “x,” “a,” “b,” and “c” represent the voltage of the system, upper asymptote, slope, and inflection point of the logistic curve, respectively.
Results and Discussion
Signal-to-noise ratio.
Optimizing the SNR is essential for obtaining the image quality needed to distinguish the various features of a sample while minimizing noise and radiation dose (Seibert 2004; Shikhaliev 2010). In this experiment, we observed a significant variability in the SNR among soils placed within different container sizes (P < 0.0001). Overall, the variation was accompanied by a consistent upward trend with an increasing voltage of the X-ray system regardless of the size of the container, as evident from the positive slopes of the linear equations fitted to various soil–container combinations (Fig. 1). The range of SNR values across all the tested containers spanned from 15 to 46 (Fig. 1). When comparing the averaged percentage increase in SNR as the X-ray system voltage progressed from 75 to 225 kV across all the tested soils, we observed a substantial increment of 37% for the 16-cm container depth. In contrast, the corresponding increase was modest, at 10%, for the 6-cm container depth. The higher SNR values observed at elevated voltages suggest the potential for enhanced image quality under such conditions. Similar findings were reported by Shikhaliev (2010), where the maximum value of SNR was achieved as the effective energy of the X-ray system increased from 5 to 42 kV while investigating adipose tissue in a 10-cm-thickness soft-tissue background. The SNR is a critical parameter for feature distinction in X-ray CT imaging. A higher SNR indicates a clearer separation between the actual signal (useful information) and background noise. In the context of soil and root studies, images with higher SNR offer improved differentiation between roots and the surrounding soil matrix. This enhanced contrast is particularly valuable in root architecture studies because it allows for more accurate segmentation and quantification of root structures. Moreover, a higher SNR can reveal finer root details and potentially capture smaller diameter roots that might be obscured in lower quality images. However, SNR alone should not be the only parameter to be considered while analyzing a sample, and it must be considered collectively with other parameters such as gray values and voltage of the system to ensure optimal dosage, enhanced image quality, and accurate interpretation (Seibert 2004; Shikhaliev 2010).
Linear relationship between signal-to-noise ratio (SNR) and X-ray computed tomography voltage (kV) for diverse soils at different bulk densities in rectangular prism containers, each 6 cm tall and of various widths (6 cm in black, 8 cm in dark yellow, 10 cm in dark blue, 12 cm in cyan, 14 cm in red, and 16 cm in green). Statistical significance is denoted by * for 5% and ** for 1%.
Citation: HortScience 59, 11; 10.21273/HORTSCI17980-24
Bulk density affected gray values.
Soil samples employed in this study (Table 1) represent various bulk densities (P < 0.0001), a key physical property conducive to a range of crop production scenarios (Jaja 2016). Although Samples A and B shared a similar texture, and similarly Samples D and E (Table 1), quantitative variations in soil particle fractions within these texture classes resulted in different bulk densities, ranging from 1.34 to 1.84 g·cm−3, a finding also supported by others (Chaudhari et al. 2013; Jones 1983). Therefore, the interpretation of this experiment is more related to bulk density than soil texture itself. Furthermore, the penetration of X-rays is more related to density than the texture of the material under consideration; an observation aligns with findings from other studies (Baveye et al. 2002; Khosravani and Reinicke 2020; Nazarian et al. 2008; Tollner and Murphy 1991). The soil Sample A, with the lowest bulk density (1.34 g·cm−3), had the highest gray value averaged across all sizes of the container whereas soil Sample E, with the highest bulk density (1.84 g·cm−3), had the lowest corresponding gray values (Supplemental Figs. 1 and 5). The gray values were significantly affected by changing voltage for a given soil sample used in this experiment (P < 0.0001). Notably, the higher bulk densities observed can be attributed to the use of air-dried soil samples before the scanning procedure. However, these values align well with simulating the upper limit of soil bulk densities for root growth in coarse-textured (1.85 g·cm−3) and fine-textured (1.47 g·cm−3) soils (Correa et al. 2019). Cultivating plants without soil moisture is unrealistic, and changes in soil moisture have a consistent effect on X-ray attenuation, allowing observations under various moisture conditions unless there are significant changes in the soil’s mineral composition (Tollner and Murphy 1991).
Estimation of voltage values ± 95% confidence intervals for plateaued gray values in X-ray computed tomography (CT) scans of soils with diverse textures, packed at various bulk densities within containers of different sizes. The fitted model parameters are given, where a, b, and c represent the upper asymptote, slope, and inflection point of the logistic curve, respectively, along with diagnostic estimates including Akaike information criterion (AIC) and root mean squared error (RMSE) used for fitting the logistic model to describe X-ray CT scan gray values against voltage.
Plateau gray value.
The logistic model adequately described X-ray penetration across tested voltages for all soil types (R2 > 0.94), providing a reasonable fit with minimal parameters (Table 1). The average observed gray values for soil Sample A with the lowest bulk density (1.34 g·cm−3) ranged from 868 to 9909, with an overall average of 4033 (Table 2 and Supplemental Fig. 1). Reduced gray values were noted in larger-sized containers, in contrast to higher gray values observed in smaller-sized containers. Comparing the percent change in gray value resulting from voltage adjustments between 75 and 225 kV, a noticeable increase of 99%, 121%, 140%, 142%, 171%, and 136% was observed for 6-, 8-, 10-, 12-, 14-, and 16-cm-sized containers, respectively (Table 2 and Supplemental Fig. 1). Regardless of the soil used in this experiment, the gray values across each container size showed an increasing trend with increasing X-ray system voltage and eventually plateauing (Table 2 and Supplemental Figures). For instance, a plateau of 9667 gray value was observed at 215 kV for 6 cm containers, 6759 at 210 kV for 8 cm, 4738 at 205 kV for 10 cm, 3628 at 230 kV for 12 cm, 2822 at 215 kV for 14 cm, and 2035 at 220 kV for 16 cm for soil Sample A (Table 1).
Average gray values of X-ray computed tomography (CT) scan with voltage variation (kV) across different sizes of containers across all tested soils packed at different bulk densities. Values within the same row followed by the same Tukey letter are not statistically different at 5%.
Similarly, for soil Sample B packed at a bulk density of 1.47 g·cm−3, the gray values exhibited variation across diverse container sizes, ranging from 784 to 9290, with an overall mean of 3545 (Table 2 and Supplemental Fig. 2). Analyzing the percentage change in gray value caused by the increase in voltage from 75 to 225 kV, distinct increments of 108%, 114%, 145%, 234%, 205%, and 107% were evident 6-, 8-, 10-, 12-, 14-, and 16-cm containers, respectively (Table 2 and Supplemental Fig. 2). A plateau of 9022 gray value was observed at 210 kV for 6-cm containers, followed by a 5524 at 190 kV for 8 cm, 4280 at 220 kV for 10 cm, 3146 at 200 kV for 12 cm, 2219 at 190 kV for 14 cm, and 1659 at 190 kV for 16 cm (Table 1).
A reduction in the observed gray values was evident when comparing soil Sample B packed at 1.47 g·cm−3 with that at 1.55 g·cm−3, mirroring higher density soil, such as soil Sample D. The soil Sample C and Sample D, with bulk densities of 1.55 g·cm−3 and 1.62 g·cm−3, respectively, exhibited similar gray values ranging from 667 to 8013 and 409 to 8083, with overall averages at 3081 and 3045 for soils Sample C and Sample D, respectively (Table 2 and Supplemental Figs. 3 and 4). Additionally, both soils, Sample C and Sample D, demonstrated similar plateau values, displaying less than a 5% difference in average gray value across corresponding voltages (Table 1). The reduced average gray values observed for soil Sample C could be attributed to the high concentration of iron oxides present in that particular soil because iron oxides have a higher effective atomic number compared with the typical components of regular soil. Consequently, X-rays were preferentially absorbed, resulting in diminished gray values, thereby highlighting the importance of accounting for the chemical composition of the soil being considered for an X-ray CT scan (Tollner and Murphy 1991).
The gray values for the most compacted soil Sample E, packed at a bulk density of 1.84 g·cm−3, exhibited a range from 592 to 6548, with an overall average of 2475 (Table 2 and Supplemental Fig. 5). Similar to prior observations, a consistent trend of initially increasing and subsequently decreasing percentage change in gray values was noted by increasing voltage from 75 to 225 kV, resulting in increments of 108%, 145%, 151%, 127%, 103%, and 101% for container sizes of 6, 8, 10, 12, 14, and 16 cm, respectively (Table 2 and Supplemental Fig. 5). Furthermore, a distinct plateau was observed, reaching 6375 at 210 kV for 6-cm containers, followed by 4284 at 225 kV for 8 cm, 2817 at 195 kV for 10 cm, 2117 at 215 kV for 12 cm, 1163 at 245 kV for 14 cm, and 1163 at 195 kV for 16 cm (Table 1). The overall decreased gray values for soil Sample E underscores the inverse correlation between soil density and the penetrative capacity of X-rays, consequently yielding reduced gray values (Nazarian et al. 2008). The results indicate that the SNR improved with increasing voltage, with gray value plateaus occurring near 200 kV. Such higher voltages exceed those commonly used in contemporary X-ray CT studies (Hou et al. 2022). The observed improvements in image quality parameters at these higher voltages suggest potential avenues for enhancing scan quality in future investigations.
Conclusions and Practical Implications
This study reveals crucial insights into optimizing X-ray CT scanning parameters for soil analysis, with significant implications for root studies. We found that increasing X-ray system voltage enhances penetration up to a plateau level, which varies depending on soil density and container size. Notably, the signal-to-noise ratio (SNR) continued to improve with increasing voltage, even as gray values plateaued around 200 kV. These finding challenges current practices in root CT protocols, which typically use voltages between 80 to 130 kV (Hou et al. 2022). Results suggest that higher voltages could potentially improve the detection of fine root structures, especially in dense soils where visibility is often compromised (Mooney et al. 2012). However, researchers must balance this potential improvement against the risk of radiation damage to living samples at higher voltages. The study also highlights the significant influence of soil mineral composition on image gray values, emphasizing the need for soil-specific calibration in CT scanning protocols. These findings provide a foundation for developing more efficient and effective CT scanning methods for soil and root research. Future work should explore a wider range of soil textures packed at various bulk densities and validate these findings with live plant specimens. Ultimately, this research contributes to advancing non-invasive plant phenotyping and soil science research, offering a pathway to more accurate and detailed studies of root architecture and soil properties.
References Cited
Araus JL, Kefauver SC, Vergara-Díaz O, Gracia-Romero A, Rezzouk FZ, Segarra J, Buchaillot ML, Chang-Espino M, Vatter T, Sanchez-Bragado R, Fernandez-Gallego JA, Serret MD, Bort J. 2022. Crop phenotyping in a context of global change: What to measure and how to do it. JIPB. 64(2):592–618. https://doi.org/10.1111/jipb.13191.
Baveye P, Rogasik H, Wendroth O, Onasch I, Crawford JW. 2002. Effect of sampling volume on the measurement of soil physical properties: Simulation with x-ray tomography data. Meas Sci Technol. 13:775–784. https://doi.org/10.1088/0957-0233/13/5/316.
Bartley PC III, Jackson BE, Fonteno WC. 2019. 3-Dimensional characterization of substrateswith X-ray microtomography. Acta Hortic. 1266:1–6. https://doi.org/10.17660/ActaHortic.2019.1266.1.
Bartley PC III. 2022. In-situ analysis of soilless substrate physical and hydraulic properties in plug production. HortScience. 57(9):S112. https://doi.org/10.21273/HORTSCI.57.9S.S1.
Brown JM, Fonteno WC, Cassel DK, Johnson GA. 1987. Computed tomographic analyses of water distribution in three porous foam media. Soil Sci Soc Am J. 51(5):1121–1125. https://doi.org/10.2136/sssaj1987.03615995005100050005x.
Chaudhari PR, Ahire DV, Ahire VD, Chkravarty M, Maity S. 2013. Soil bulk density as related to soil texture, organic matter content and available total nutrients of coimbatore soil. Int J Sci Res Pub. 3(2):1–8.
Correa J, Postma JA, Watt M, Wojciechowski T. 2019. Soil compaction and the architectural plasticity of root systems. J Exp Bot. 70(21):6019–6034. https://doi.org/10.1093/jxb/erz383.
Flavel RJ, Guppy CN, Rabbi SMR, Young IM. 2017. An image processing and analysis tool for identifying and analysing complex plant root systems in 3D soil using non-destructive analysis: Root1. PLoS One. 12(5):e0176433. https://doi.org/10.1371/journal.pone.0176433.
Gao W, Schlüter S, Blaser SRGA, Shen J, Vetterlein D. 2019. A shape-based method for automatic and rapid segmentation of roots in soil from X-ray computed tomography images: Rootine. Plant Soil. 441(1–2):643–655. https://doi.org/10.1007/s11104-019-04053-6.
Gregory PJ, George TS. 2011. Feeding nine billion: The challenge to sustainable crop production. J Exp Bot. 62(15):5233–5239. https://doi.org/10.1093/jxb/err232.
Hou L (Helen), Gao W, van der Bom F, Weng Z (Han), Doolette CL, Maksimenko A, Hausermann D, Zheng Y, Tang C, Lombi E, Kopittke PM. 2022. Use of X-ray tomography for examining root architecture in soils. Geoderma. 405:115405. https://doi.org/10.1016/j.geoderma.2021.115405.
Jaja N. 2016. Understanding the texture of your soil for agricultural productivity. Virginia Cooperative Extension Publication CSES-162P.
Jones CA. 1983. Effect of soil texture on critical bulk densities for root growth. Soil Sci Soc Am J. 47(6):1208–1211. https://doi.org/10.2136/sssaj1983.03615995004700060029x.
Khosravani MR, Reinicke T. 2020. On the use of x-ray computed tomography in assessment of 3d-printed components. J Nondestruct Eval. 39(4). https://doi.org/10.1007/s10921-020-00721-1.
Kirk GJD, Boghi A, Affholder MC, Keyes SD, Heppell J, Roose T. 2019. Soil carbon dioxide venting through rice roots. Plant Cell Environ. 42(12):3197–3207. https://doi.org/10.1111/pce.13638.
Krbcova Z, Kukal J. 2017. Relationship between entropy and SNR changes in image enhancement. J Image Video Proc. 2017(1). https://doi.org/10.1186/s13640-017-0232-z.
Li A, Zhu L, Xu W, Liu L, Teng G. 2022. Recent advances in methods for in situ root phenotyping. PeerJ. 10:e13638. https://doi.org/10.7717/peerj.13638.
Lynch JP. 2022. Harnessing root architecture to address global challenges. Plant J. 109(2):415–431. https://doi.org/10.1111/tpj.15560.
Lynch JP, Mooney SJ, Strock CF, Schneider HM. 2021. Future roots for future soils. Plant Cell Environ. 45(3):620–636. https://doi.org/10.1111/pce.14213.
Mooney SJ, Pridmore TP, Helliwell J, Bennett MJ. 2012. Developing X-ray computed tomography to non-invasively image 3-D root systems architecture in soil. Plant Soil. 352(1–2):1–22. https://doi.org/10.1007/s11104-011-1039-9.
Nazarian A, Snyder BD, Zurakowski D, Müller R. 2008. Quantitative micro-computed tomography: A non-invasive method to assess equivalent bone mineral density. Bone. 43(2):302–311. https://doi.org/10.1111/tpj.15560.
Pires LF, Borges JAR, Bacchi OOS, Reichardt K. 2010. Twenty-five years of computed tomography in soil physics: A literature review of the Brazilian contribution. Soil Tillage Res. 110(2):197–210. https://doi.org/10.1016/j.still.2010.07.013.
Rana N, Rawat D, Parmar M, Dhawan D, Bhati A, Mittal B. 2015. Evaluation of external beam hardening filters on image quality of computed tomography and single photon emission computed tomography/computed tomography. J Med Phys. 40(4):198–206. https://doi.org/10.4103/0971-6203.170790.
Seibert JA. 2004. Tradeoffs between image quality and dose. Pediatr Radiol. 34:S183. https://doi.org/10.1007/s00247-004-1268-7.
Shikhaliev PM. 2010. The upper limits of the SNR in radiography and CT with polyenergetic X-rays. Phys Med Biol. 55(18):5317–5339. https://doi.org/10.1088/0031-9155/55/18/005.
Shikhaliev PM. 2005. Beam hardening artifacts in computed tomography with photon counting, charge integrating and energy weighting detectors: A simulation study. Phys Med Biol. 50(24):5813–5827. https://doi.org/10.1088/0031-9155/50/24/004.
Sprawls P. 1995. Physical principles of medical imaging. Medical Physics Pub, Madison, WI, USA.
Stock SR. 2008. Recent advances in X-ray microtomography applied to materials. Int Mat Rev. 53(3):129–181. https://doi.org/10.1179/174328008X277803.
Tollner EW, Murphy C. 1991. Factors affecting soil X-ray absorption coefficients with computed tomography. Trans ASAE. 34(3):1047–1053.
Wickham H. 2016. ggplot2: Elegant graphics for data analysis. Springer-Verlag.
Wildenschild D, Hopmans JW, Vaz CMP, Rivers ML, Rikard D, Christensen BSB. 2002. Using X-ray computed tomography in hydrology: Systems, resolutions and limitations. J Hydrol. 267(3–4):285–297. https://doi.org/10.1016/S0022-1694(02)00157-9.
Yafuso EJ, Fisher PR, Bohórquez AC. 2019. Quantification of poinsettia root growth by image scans or X-ray computed tomography scans in three propagation substrates at varied moisture levels. Comm Soil Sci Plant Anal. 50(18):2354–2367. https://doi.org/10.1080/00103624.2019.1659305.
Young DS. 2017. Handbook of regression methods. CRC Press, Boca Raton, FL, USA.