Paun et al. European Radiology Experimental (2020) 4:33 https://doi.org/10.1186/s41747-020-00163-4 European Radiology Experimental ORIGINAL ARTICLE Open Access Modelling the skeletal muscle injury recovery using in vivo contrast-enhanced micro-CT: a proof-of-concept study in a rat model Bruno Paun1, Daniel García Leon1, Alex Claveria Cabello1, Roso Mares Pages1, Elena de la Calle Vargas1, Paola Contreras Muñoz2,3, Vanessa Venegas Garcia2,3, Joan Castell-Conesa1, Mario Marotta Baleriola2,3 and Jose Raul Herance Camacho1* Abstract Background: Skeletal muscle injury characterisation during healing supports trauma prognosis. Given the potential interest of computed tomography (CT) in muscle diseases and lack of in vivo CT methodology to image skeletal muscle wound healing, we tracked skeletal muscle injury recovery using in vivo micro-CT in a rat model to obtain a predictive model. Methods: Skeletal muscle injury was performed in 23 rats. Twenty animals were sorted into five groups to image lesion recovery at 2, 4, 7, 10, or 14 days after injury using contrast-enhanced micro-CT. Injury volumes were quantified using a semiautomatic image processing, and these values were used to build a prediction model. The remaining 3 rats were imaged at all monitoring time points as validation. Predictions were compared with Bland- Altman analysis. Results: Optimal contrast agent dose was found to be 20 mL/kg injected at 400 μL/min. Injury volumes showed a decreasing tendency from day 0 (32.3 ± 12.0mm3, mean ± standard deviation) to day 2, 4, 7, 10, and 14 after injury (19.6 ± 12.6, 11.0 ± 6.7, 8.2 ± 7.7, 5.7 ± 3.9, and 4.5 ± 4.8 mm3, respectively). Groups with single monitoring time point did not yield significant differences with the validation group lesions. Further exponential model training with single follow-up data (R2 = 0.968) to predict injury recovery in the validation cohort gave a predictions root mean squared error of 6.8 ± 5.4 mm3. Further prediction analysis yielded a bias of 2.327. Conclusion: Contrast-enhanced CT allowed in vivo tracking of skeletal muscle injury recovery in rat. Keywords: Muscle (skeletal), Muscular diseases, Rats, Tomography (x-ray computed), Wound healing * Correspondence: raul.herance@vhir.org 1 Medical Molecular Imaging Group, Vall d’Hebron Research Institute (VHIR), CIBER-BBN, CIBBIM-Nanomedicine, ISCIII, Hospital Universitari Vall d’Hebron, Universitat Autònoma de Barcelona (UAB), Passeig de la Vall d’Hebron 119-129, 08035 Barcelona, Spain Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Paun et al. European Radiology Experimental (2020) 4:33 Page 2 of 8 Key points where the authors use ex vivo contrast-enhanced micro- CT to quantify underlying angiogenesis in the process of  Contrast-enhanced micro-computed tomography wound healing. Therefore, the lack of methodologies showed applicability in monitoring skeletal muscle allowing in vivo assessment of SM recovery was the injury recovery. main motivation of the present work.  Highest accuracy in lesion recovery predictions was Consequently, we aimed to (i) present a new method- obtained on day 14 after injury. ology to image and quantify in vivo SM recovery by  We provided a proof of concept to track skeletal contrast-enhanced micro-CT means and (ii) propose a muscle healing stages using CT-derived volume- proof-of-principle study of a SM recovery prediction based lesion recovery. model. Background Methods Skeletal muscle (SM) represents about 40% of the body Animal preparation mass and is formed by contractile multinucleated muscle The study was approved by the Local Animal Ethics fibres as a result from myoblasts fusion. Injuries in SM Committee and was performed in accordance with may come from diverse events, including direct trauma Spanish (RD 53/2013) and European (2010/63/UE) le- from muscle contusions, lacerations and strains or indir- gislation. Wistar male adult rats (n = 27), weighing ect trauma from degenerative diseases like muscular dys- approximately 425 g each (425.4 ± 43 g), were housed trophy, among others. Professional athletes are the most at 22–24 °C, maintained on a 12-h light/dark cycle, affected by SM injuries, representing from 10 to 55% of and used for the experiments. Water and food were all injuries in this population [1, 2]. Moreover, the type, given ad libitum during the experiments. Anaesthesia severity, size and location of the injury make inference was induced with 4% isoflurane (Aerrane, Baxter of recovery and rehabilitation times challenging, thus Healthcare, Deerfield, IL, USA) and maintained with directly affecting prognosis of lay-off times [3]. 1.5–2.0% isoflurane in O2 at 1 L/min during the In mammals, SM has the ability to regenerate in re- whole study. A 22-gauge intravenous cannula (Intro- sponse to injury [4]. This process can be divided into can Safety® IV Catheter, 22G, B. Braun, Melsungen, three phases [1, 4–6]: (1) degeneration, characterised Germany) was placed in the tail vein for iopamidol by muscle fibres rupture and necrosis, formation of (iopamiro 370 mg/mL, Bracco Imaging, Milan, Italy) hematoma and inflammatory cell reaction; (2) regener- administration as contrast agent (CA) by means of an ation, characterised by phagocytosis of the damaged infusion pump (Harvard apparatus PHD 2000 Infu- tissue, satellite cells (SM stem cells) activation and sion, Holliston, MA, USA). Post-surgically, 250 mL of angiogenesis; and (3) remodelling, characterised by the water with 10% of buprenorphine (buprecare 0.3 mg/ maturation and reorganisation of regenerated muscle mL, Divasa Farmavic, Barcelona, Spain) was given to fibres, recovery of muscle function together with fi- all operated animals. brosis and scar tissue formation. The new vascular network emerges from surviving trunks of the blood vessels into the injured region providing means for Contrast agent dose optimisation oxygen delivery and nutrients necessary to recover Optimal CA dose to distinguish between injury and mature myofibres [1, 7]. neighbouring tissue was first determined by performing To characterise in vivo muscle injury, both magnetic a contrast enhancement study in a separate set of experi- resonance imaging (MRI) and ultrasound are well- ments including five rats. These were placed in the established imaging approaches in hospitals [8]. How- micro-CT bed, and five different infusion rates (100, ever, there is a knowledge gap regarding the approaches 200, 300, 400, and 500 μL/min) were tested for 20 min to to monitor and quantify SM injury using computed tom- obtain different contrast-to-noise ratios (CNR) in ac- ography (CT), which plays an important role in tissue quired images. Imaging parameters were as follows: field density characterisation in muscle diseases [9]. Along of view 40 mm, voxel size 80 μm (isotropic), kVp 90, with its application in whole-body scans, CT is also part tube current 200 μA and exposure time 120 s. Images of hybrid scanners as single-photon computed tomog- were acquired approximately every 2.5 min, resulting in raphy/CT or positron emission tomography (PET)/CT total of 9 images per rat. Contrast enhancement was where functional studies can be conducted to monitor assessed in the femoral artery and neighbouring tissue SM lesion healing. by delineating circular volumes of interest of 1 mm in To the best of our knowledge, there are no studies diameter, obtaining their mean density and standard de- where CT has been used for in vivo tracking of SM in- viation values. From these, we calculated CNR using fol- jury. There are several studies in the literature [7, 10], lowing formula [11]: Paun et al. European Radiology Experimental (2020) 4:33 Page 3 of 8 jμA −μT j were imaged at all mentioned follow-up days as valid- CNR ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi; ation. High-resolution images were acquired so that σ 2A þ σ 2T the centre of the field of view was aligned with the where μA and μT represent mean density values of vol- middle of the fibula in sagittal and coronal planes, umes of interest in femoral artery and tissue area, re- thus covering the maximal region of the left limb spectively, and σA and σT represent their standard containing the injury. Experimental imaging parame- deviations. Optimal infusion rate providing adequate ters were field of view 40 mm, voxel size 80 μm (iso- image contrast was set following the Rose criterion tropic), kVp 90, tube current 200 μA and exposure (CNR > 5), as previously described [12]. time 270 s. All images were reconstructed using a fil- After CA administration, the pump was stopped, and tered back-projection approach with a Ram-Lak filter, micro-CT images were acquired. One millilitre of saline including a ring reduction algorithm [13]. solution was injected after image acquisition was finished. Image processing Greyscale values between image datasets were normal- Skeletal muscle injury model ised by an histogram matching algorithm implemented For the injury model, we used a recently developed sur- in the Insight Toolkit [14]. All image datasets in the gically induced SM injury rat model proposed by study were normalised by matching their histograms to Contreras-Muñoz et al [6]. Briefly, 23 anaesthetised rats the reference image histogram, which was chosen to be were immobilised by the fixation of tail and extremities the injury image of rat 1 at day 0. Adapted normalised with adhesive strips to a styrofoam surface exposing the image datasets were then filtered in Fiji [15] using a ventral side of the right crus. Skeletal traumatic muscle non-local means denoising filter [16, 17]. Noise standard injuries were induced in the rat medial gastrocnemius deviation was set to 2 with smoothing factor set to 1. muscle by a 18-gauge biopsy needle (Bard® Monopty® Contrast-enhanced anatomy, together with high- Disposable Core Biopsy Instrument, Bard Biopsy Sys- intensity anatomy like bones, was segmented from tems, Tempe, USA) with a 0.84-mm inner diameter. denoised images using thresholding based on Renyi en- Transversal biopsy procedure was performed at the tropy [18] with min and max values set to 77 and 255, muscle-tendon junction level of the left leg medial respectively. These values were experimentally derived gastrocnemius muscle (3 mm from the start of muscle- from denoised images to ease the injury segmentation tendon junction and 2 mm in depth). Immediately after process. All segmented contrast-enhanced anatomy im- muscle injury, a cannula (Introcan Safety® IV Catheter, ages were then manually processed in three-dimensional 22G, B. Braun, Melsungen, Germany) was introduced slicer (version 4.9.0) [19]. The injury was further seg- within the injury in order to proceed with the micro-CT mented from contrast-enhanced anatomy mask, exclud- imaging protocol. ing surrounding adipose tissue and vasculature. Once the refined SM injury mask was obtained, injury volume Micro-CT was calculated as the mask voxel number multiplied by Micro-CT studies were performed using a Quantum FX voxel volume. micro-CT scanner (PerkinElmer, Hopkinton, MA, USA). Rats (n = 23) were positioned on their right side in a Lesion recovery prediction model bed, and their left leg was immobilised to minimise pos- Single follow-up day cohort data (n = 20) were used to sible involuntary movement and motion-related arte- feed an exponential model of SM injury recovery. The facts. During the scans, rats were kept under generalised exponential model used was Vt = V0 × exp anaesthesia, and CA was administered for 20 min as de- (b × t), where t is the corresponding post-injury time scribed above. Two images were acquired at the same (days), V0 is the initial injury size (mm3) and b is the day when muscle injury was induced: first an image of healing rate (days−1). Model was trained using the least the injury with the cannula at 14 min after the beginning absolute residual method available in Matlab (The Math- of CA infusion for location purposes, and then, after re- works Inc., Natick, MA, USA), quantifying both R2 and moving the cannula from the injury, a second image was root mean squared error. Afterwards, injury volumes acquired at 20 min right after stopping the CA injection. were predicted in the validation cohort (n = 3) at all Again, 1 mL of saline solution was injected after image monitored post-injury time points. acquisition was finished to ease contrast clearance. To image lesion recovery at different time points, Statistical analysis 20 rats were sorted into 5 different groups (n = 4 per Data are expressed as mean and standard deviation. In- group) according to the follow-up day at 2, 4, 7, 10, jury volumes and model predictions were compared or 14 days after injury, respectively. Additional 3 rats using paired t test and Bland-Altman analysis after Paun et al. European Radiology Experimental (2020) 4:33 Page 4 of 8 testing for normality. All statistical analyses were per- between single time point groups and validation group formed using the Graphpad Prism software (Graphpad at day 0 (34.0 ± 11.9 versus 21.4 ± 3.8 mm3, p = 0.088), Software Inc., San Diego, CA, USA). A p value below day 2 (22.1 ± 14.5 versus 16.3 ± 11.7 mm3, p = 0.598), 0.05 was considered as statistically significant. day 4 (11.3 ± 5.3 versus 10.7 ± 9.6 mm3, p = 0.915), day 7 (9.2 ± 9.7 versus 7.0 ± 5.9 mm3, p = 0.744), day 10 (6.4 Results ± 4.5 versus 4.7 ± 3.8 mm3, p = 0.622) and at day 14 (3.5 Optimal contrast agent dose ± 3.0 versus 5.8 ± 0.4 mm3, p = 0.584), respectively. Prior to tracking injury healing by micro-CT, analysis of the influence of CA dose in image contrast was per- Injury volume predictions formed to determine optimal infusion rate. Figure 1 Model training with single follow-up data yielded a cor- shows obtained CNR curves for each tested infusion relation of R2 = 0.968, with root-mean-square error = rate. Since cutoff criterion was set at CNR over 5 after 1.77 mm3. This was further tested with the validation co- 20 min of infusion, contrast injection rates of 400 and hort (n = 3) at all monitored time points (days 2, 4, 7, 500 μL/min were the only ones satisfying this require- 10, and 14 after injury). Mean root-mean-square error in ment. However, to reduce the amount of CA injected to predictions was 6.8 ± 5.4 mm3. As presented in the animals and to avoid possible side effects, 400 μL/min Bland-Altman plot (Fig. 4), overall predictions yielded a was chosen as optimal rate and was used in following bias of 2.3 (limits of agreement at 95%, -13.5 and 18.1). study, resulting in total administered amount of contrast The exponential model better performed at low average of 20 mL/kg per rat. values, offering higher accuracy for nearly healed injury time points. In vivo SM lesion tracking In Fig. 2, one example of SM lesion healing in short-axis Discussion view monitored at all post-injury time points is pre- In this proof-of-concept study, we assessed the in vivo sented. As shown, lesion healing process remains rather CT applicability to track the progression of SM injury active up to 2 weeks after injury. This was further quan- recovery. For this purpose, we previously defined a con- tified in Fig. 3, sorting lesion volumes according to the trast enhancement CT protocol in order to guarantee day injury was imaged. The mean initial lesion size (n = proper differentiation between injury and healthy tissue. 23) was 32.3 ± 12.0 mm3. As expected, lesion healing After quantifying injury volumes during follow-up to process yielded decreasing injury volumes (n = 7) at day track injury recovery, an exponential prediction model 2 (19.6 ± 12.6 mm3), day 4 (11.0 ± 6.7 mm3), day 7 (8.2 of injury recovery has been proposed to mimic the ob- ± 7.7 mm3), day 10 (5.7 ± 3.9 mm3), and day 14 (4.47 ± served decreasing tendency during healing. Although 4.8 mm3) post-injury. Additionally, groups with a single simple, the model offered a higher accuracy for late monitored time point after injury (5 time points, n = 4 follow-up time points than that for early time points each) were compared to the validation cohort (n = 3) to when the lesion is nearly recovered. support this decreasing tendency. Indeed, lesion volumes Regarding the contrast enhancement protocol, we did not show any statistically significant differences found a CA volume of 20 mL/kg that was necessary to Fig. 1 Contrast-to-noise ratio quantification for several infusion rates. Tested contrast injection rates from 100 to 500 μL/min (yellow), with an overall administration time of 20 min. The red dashed line defines the applied cutoff criterion to determine the optimal infusion rate. CNR Contrast-to-noise ratio Paun et al. European Radiology Experimental (2020) 4:33 Page 5 of 8 Fig. 2 Contrast-enhanced micro-computed tomography images and segmented injuries from one representative example of the validation cohort. a First row shows short-axis images of the lesion in similar location during all follow-up time points. Images were reconstructed using a filtered back-projection approach with a Ram-Lak filter. Second row images superimpose the obtained segmentation mask (blue) of the injury on the first row images. b Three-dimensional volumetric rendering of the segmented lesions (white) during all follow-up time points obtain a proper CNR on CT images. Injecting more con- previously reported, this injury model well reproduces trast may cause unnecessary side effects in rats without the human muscle strain injuries observed in profes- significant increasing image contrast. In addition, the sional athletes. This was characterised by means of hist- proposed injection rate of 400 μL/min was found to be ology and MRI (standard of care in muscle lesion the best rate allowing a proper biodistribution of the CA tracking), showing the healing capacity observed in ani- without causing any side effect in rats. This experimen- mal models of spontaneous skeletal muscle injury as the tally determined CA amount was in agreement with dystrophic mdx-mice [21]. Although similar looking guidelines about intravenous administration volumes for skeletal muscle injuries on MRI at the same anatomical rats reported by Diehl et al [20]. site, different healing rates have been observed as the le- The murine model of skeletal muscle injury used in sion nature varies (either high-speed running or over- this study has been previously described in detail [6]. As stretching) in athletes [22]. This study focused on the applicability of in vivo CT imaging to track SM lesion recovery. However, estab- lished imaging techniques for prognosis and muscle re- pair monitoring in the clinics are both MRI and ultrasound [8]. The latter is normally considered to be first choice for clinical diagnosis due to its low price and fast applicability, but MRI is preferred when detailed characterisation of injury is required [23]. Recently, Todeschini et al. [24] reported a series of 39 professional soccer players to study pubalgia imaging features with both ultrasound and MRI. Interestingly, from 9 individ- uals with a positive MRI scan regarding lesions of the common aponeurosis of the rectus abdominis/adductor longus muscles, only 4 yielded a positive ultrasound scan. MRI has been positioned as a widely used tech- Fig. 3 In vivo three-dimensional skeletal muscle injury volume nique in muscle trauma evaluation, playing an increas- quantification up to 14 days after injury. This quantification includes ingly role in grading injury severity and guiding return both single follow-up (n = 20) and validation (n = 3) cohorts to play in the injured athlete [25]. Nonetheless, this Paun et al. European Radiology Experimental (2020) 4:33 Page 6 of 8 Fig. 4 Bland-Altman analysis of the predicted injury volumes in the validation cohort. Predictions were compared to the injury volumes during all follow-up time points. Red dotted line corresponds to the bias, with corresponding lower and upper limits of agreement at 95% represented with green dotted lines technique does not offer any information regarding tis- automatically learn hierarchical features from raw im- sue dynamics when healing as a counterpart. Leaning on ages to classify these according to a training classifica- hybrid approaches using PET /CT or PET/MRI scans, tion [29]. For instance, Zhang and colleagues [30] this may be of relevance to study SM lesion healing dy- presented a CNN model to classify soft tissue sarcoma namics in athletes. grading using both CT and MR images from a series of There are just a few studies using PET/CT to assess 51 patients, offering great performance in this task. In muscle injury. For instance, Carter et al [26] studied this matter, CNNs could definitely be applied to track metabolic changes related to thermally induced muscle skeletal muscle injury healing, supporting current injury in rabbits using 2-deoxy-2-[fluorine-18]fluoro-D- trauma prognosis. glucose (18F-fluorodeoxyglucose, 18F-FDG) PET. Xie This study has several limitations. First, there is a lim- et al [27] showed that mice with muscle injury induced ited number of rats included in the validation cohort. by electroporation show increased uptake of 64Cu, thus Although tracking full progression of the injury recovery demonstrating 64Cu-Cl2 PET/CT as a new molecular im- in all animals would have been desirable, no statistical aging technique for skeletal muscle. Recently, Pervaiz differences were observed between the single follow-up et al [28] showed that 18F-FDG PET/CT can be used for day and the validation cohorts. Second, no histological imaging inflammation and muscle injury caused by chol- analysis of the lesions was provided. Conversely, re- esterol crystal emboli. However, these studies focused on ported three-dimensional SM lesion characterisation the use of radiopharmaceuticals to assess a specific data are consistent with previously reported histological physiological disorder that cannot explain the whole SM results. Regarding lesion recovery, a direct association to recovery. Merging these findings with in vivo contrast- professional athlete lesions cannot be established as the enhanced CT quantifying muscle recovery, both imaging lesion nature varies from the surgically induced one in techniques may offer enriched lesion healing structural the rat model. Also, in vivo SM lesion tracking cannot and functional information, yielding a more accurate in- be directly extrapolated to human studies. Indeed, the jury prognosis, among other uses. amount of iodine-based contrast agent used in our ex- Although simple, this study also presented an expo- periments exceeds the maximum recommended dose for nential model of muscle recovery as a proof of concept. contrast enhancement studies included in current clin- This model relied on quantified injury volumes through ical guidelines. To overcome the low soft tissue contrast time to mimic the observed decreasing tendency derived problem during SM injury monitoring, early injury from two-dimensional histological analyses in previous stages may offer a good starting point for tracking. For studies [6]. Yielded predictions were accurate when le- instance, both oedema and inflammatory processes in sion volumes were nearly recovered but differed more at clinical contrast-enhanced CT images could serve as ini- few days after injury. One way to better predict lesion tial landmarks to ease injury segmentation process. recovery at all monitored time points could be to detect Adapting this prior segmentation to subsequent post- additional lesion features via convolutional neural net- injury examinations leaning on anatomical references work (CNN) models. These deep learning models can detection, recovery predictions and the expert Paun et al. European Radiology Experimental (2020) 4:33 Page 7 of 8 knowledge to refine lesion border delineation accord- Received: 8 November 2019 Accepted: 22 April 2020 ingly may offer similar outcomes as the ones we presented from animal experiments. Of note, acquired z- axis resolution in conventional CT scanners (~ 3 mm) References might hinder this task, so detailed studies would be re- 1. Järvinen TA, Järvinen TL, Kääriäinen M, Kalimo H, Järvinen M (2005) Muscle injuries: biology and treatment. Am J Sports Med 33:745–764 https://doi. quired to minimise the use of CA. Altogether, additional org/10.1177/0363546505274714 in vivo lesion tracking using contrast-enhanced CT in 2. Garg K, Corona BT, Walters TJ (2015) Therapeutic strategies for preventing large animal models first, and then in patient studies, skeletal muscle fibrosis after injury. Front Pharmacol 6. https://doi.org/10. would be required to further validate these results. 3389/fphar.2015.00087 3. Ekstrand J, Healy JC, Waldén M, Lee JC, English B, Hägglund M (2012) In conclusion, our study showed that contrast- Hamstring muscle injuries in professional football: the correlation of MRI enhanced CT can allow in vivo tracking of SM healing findings with return to play. Br J Sports Med 46:112–117 https://doi.org/10. after injury, presenting a proof of concept of an expo- 1136/bjsports-2011-090155 4. Laumonier T, Menetrey J (2016) Muscle injuries and strategies for improving nential recovery model, in a rat model. their repair. J Exp Orthop 3 https://doi.org/10.1186/s40634-016-0051-7 5. Chargé SBP, Rudnicki MA (2004) Cellular and molecular regulation of muscle Abbreviations regeneration. Physiol Rev 84:209–238 https://doi.org/10.1152/physrev.00019. CA: Contrast agent; CNN: Convolutional neural network; CNR: Contrast-to- 2003 noise ratio; CT: Computed tomography; FDG: Fluorodeoxyglucose; 6. Contreras-Muñoz P, Fernández-Martín A, Torrella R et al (2016) A new MRI: Magnetic resonance imaging; PET: Positron emission tomography; surgical model of skeletal muscle injuries in rats reproduces human sports SM: Skeletal muscle lesions. Int J Sports Med 37:183–190 https://doi.org/10.1055/s-0035-1555933 7. Singh DP, Barani Lonbani Z, Woodruff MA, Parker TJ, Steck R, Peake JM Authors’ contributions (2017) Effects of topical icing on inflammation, angiogenesis, Conceptualisation and formulation of the ideas presented in this manuscript revascularization, and myofiber regeneration in skeletal muscle following were done by JRHC, BP, and ACC. Data curation and statistical, mathematical, contusion injury. Front Physiol 8:1–15 https://doi.org/10.3389/fphys.2017. computational, or other formal analysis of the data were performed by BP 00093 and ECV. Acquisition of the financial support for the project leading to this 8. Slavotinek JP (2010) Muscle injury: the role of imaging in prognostic publication was done by JRHC, JCC, and MMB. Research and investigation assignment and monitoring of muscle repair. Semin Musculoskelet Radiol process, more precisely experiments and data/evidence collection were 14:194–200 https://doi.org/10.1055/s-0030-1253160 performed by BP, ACC, RMP, VVG, PCM, JCC, and MMB. Development and 9. Engelke K, Museyko O, Wang L, Laredo JD (2018) Quantitative analysis of design of methodology and creation of models were done by BP, PCM, and skeletal muscle by computed tomography imaging—state of the art. J JRHC. Provision of study materials, reagents, materials, patients, laboratory Orthop Transl 15:91–103 https://doi.org/10.1016/j.jot.2018.10.004 samples, animals, instrumentation, computing resources, or other analysis 10. Urao N, Okonkwo UA, Fang MM, Zhuang ZW, Koh TJ, DiPietro LA (2016) tools was performed by JRHC, ACC, RMP, and VVG. Programming, MicroCT angiography detects vascular formation and regression in skin implementation of the computer code and supporting algorithms and wound healing. Microvasc Res 106:57–66 https://doi.org/10.1016/j.physbeh. testing of existing code components were performed by BP and ECV. 2017.03.040 Validation of the results/experiments and other research outputs was done 11. Demirturk Kocasarac H, Helvacioglu Yigit D, Bechara B, Sinanoglu A, by JRHC, BP. JRHC, BP, ACC, MMB, and JCC contributed in writing, reviewing Noujeim M (2016) Contrast-to-noise ratio with different settings in a CBCT and editing of the manuscript. All authors read and approved the final machine in presence of different root-end filling materials: an in vitro study. manuscript. Dentomaxillofacial Radiol 45. https://doi.org/10.1259/dmfr.20160012 12. Burgess AE (1999) The Rose model, revisited. J Opt Soc Am A Opt Image Sci Funding Vis 16:633 https://doi.org/10.1364/josaa.16.000633 This study has received funding by European Commission (Horizon 2020), 13. Jang SY, Kim HK, Youn H, Cho S, Cunningham IA (2017) Fourier analysis of the Carlos III Health Institute and the European Regional Development Fund noise characteristics in cone-beam microtomography laboratory scanners. (PI16/02064) and AGAUR from Generalitat de Catalunya (2017SGR1303). IEEE Trans Biomed Eng 64:173–183 https://doi.org/10.1109/TBME.2016. 2552496 Availability of data and materials 14. Yoo TS, Ackerman MJ, Lorensen WE et al (2002) Engineering and algorithm The datasets used and/or analysed during the current study are available design for an image processing API: a technical report on ITK - The Insight from the corresponding author on reasonable request. Toolkit. In: Studies in Health Technology and Informatics. IOS Press, pp 586– 592 https://doi.org/10.3233/978-1-60750-929-5-586 Ethics approval 15. Schindelin J, Arganda-Carreras I, Frise E et al (2012) Fiji: An open-source Approval from the institutional animal care committee was obtained. platform for biological-image analysis. Nat Methods 9:676–682 https://doi. org/10.1038/nmeth.2019 Consent for publication 16. Buades A, Coll B, Morel J-M (2011) Non-local means denoising. Image Not applicable. Processing On Line 1:208–212 https://doi.org/10.5201/ipol.2011.bcm_nlm 17. Darbon J, Cunha A, Chan TF, Osher S, Jensen GJ (2008) Fast nonlocal Competing interests filtering applied to electron cryomicroscopy. IEEE:1331–1334 https://doi.org/ The authors declare that they have no competing interests. 10.1109/ISBI.2008.4541250 18. Wang S, Jia H, Peng X (2020) Modified salp swarm algorithm based Author details multilevel thresholding for color image segmentation. Math Biosci Eng 17: 1 Medical Molecular Imaging Group, Vall d’Hebron Research Institute (VHIR), 700–724 https://doi.org/10.3934/mbe.2020036 CIBER-BBN, CIBBIM-Nanomedicine, ISCIII, Hospital Universitari Vall d’Hebron, 19. Fedorov A, Beichel R, Kalpathy-Cramer J et al (2012) 3D Slicer as an image Universitat Autònoma de Barcelona (UAB), Passeig de la Vall d’Hebron computing platform for the quantitative imaging network. Magn Reson 119-129, 08035 Barcelona, Spain. 2Health & Biomedicine division, Leitat Imaging 30:1323–1341 https://doi.org/10.1016/j.mri.2012.05.001 Technological Center, 2. C/ Pallars, 179-185, 08005 Barcelona, Spain. 20. Diehl K, Hull R, Morton D et al (2001) A good practice guide to the 3 Bioengineering, Cell therapy and Surgery in Congenital Malformations administration of substances and removal of blood, including routes and Laboratory, Vall d’Hebron Research Institute (VHIR), Hospital Universitari Vall volumes. J Appl Toxicol 23:15–23 https://doi.org/10.1002/jat.727 d’Hebron, Universitat Autònoma de Barcelona (UAB), Passeig de la Vall 21. Marotta M, Sarria Y, Ruiz-Roig C, Munell F, Roig-Quilis M (2007) Laser d’Hebron 119-129, 08035 Barcelona, Spain. microdissection-based expression analysis of key genes involved in muscle Paun et al. European Radiology Experimental (2020) 4:33 Page 8 of 8 regeneration in mdx mice. Neuromuscul Disord 17:707–718 https://doi.org/ 10.1016/j.nmd.2007.05.007 22. Askling CM, Malliaropoulos N, Karlsson J (2012) High-speed running type or stretching-type of hamstring injuries makes a difference to treatment and prognosis. Br J Sports Med 46:86–87 https://doi.org/10.1136/bjsports-2011- 090534 23. Järvinen TAH, Järvinen TLN, Kääriäinen M et al (2007) Muscle injuries: optimising recovery. Best Pract Res Clin Rheumatol 21:317–331 https://doi. org/10.1016/j.berh.2006.12.004 24. Todeschini K, Daruge P, Bordalo-Rodrigues M, Pedrinelli A, Busetto AM (2019) Imaging assessment of the pubis in soccer players. Rev Bras Ortop 54:118–127 https://doi.org/10.1016/j.rbo.2017.12.012 25. Flores DV, Mejía Gómez C, Estrada-Castrillón M, Smitaman E, Pathria MN (2018) MR imaging of muscle trauma: anatomy, biomechanics, pathophysiology, and imaging appearance. Radiographics 38:124–148 https://doi.org/10.1148/rg.2018170072 26. Carter EA, Tompkins RG, Hsu H et al (1997) Metabolic alterations in muscle of thermally injured rabbits, measured by positron emission tomography. Life Sci 61:39–44 https://doi.org/10.1016/S0024-3205(97)00355-X 27. Xie F, Cai H, Peng F (2017) CuCl 2 PET / CT imaging of mouse muscular injury induced by electroporation. Am J Nucl Med Mol Imaging 7:33–39 Retrieved from http://www.ajnmmi.us/ 28. Pervaiz MH, Durga S, Janoudi A, Berger K, Abela GS (2018) PET/CTA detection of muscle inflammation related to cholesterol crystal emboli without arterial obstruction. J Nucl Cardiol 25:433–440 https://doi.org/10. 1007/s12350-017-0826-y 29. Huang L, Gong ZY, Jun YT (2019) Skin lesion segmentation using object scale-oriented fully convolutional neural networks. Signal Image Video Process 13:431–438 https://doi.org/10.1007/s11760-018-01410-3 30. Zhang L, Ren Z (2019) Comparison of CT and MRI images for the prediction of soft-tissue sarcoma grading and lung metastasis via a convolutional neural networks model. Clin Radiol 75:64–69 https://doi.org/10.1016/j.crad. 2019.08.008 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.