Background: The purpose of this study was to investigate whether any texture features show a correlation with intrahepatic tumor growth before the metastasis is visible to the human eye. Methods: Eight male C57BL6 mice (age 8–10 weeks) were injected intraportally with syngeneic MC-38 colon cancer cells and two mice were injected with phosphate-buffered saline (sham controls). Small animal magnetic resonance imaging (MRI) at 4.7 T was performed at baseline and days 4, 8, 12, 16, and 20 after injection applying a T2-weighted spin-echo sequence. Texture analysis was performed on the images yielding 32 texture features derived from histogram, gray-level co-occurrence matrix, gray-level run-length matrix, and gray-level size-zone matrix. The features were examined with a linear regression model/Pearson correlation test and hierarchical cluster analysis. From each cluster, the feature with the lowest variance was selected. Results: Tumors were visible on MRI after 20 days. Eighteen features from histogram and the gray-level-matrices exhibited statistically significant correlations before day 20 in the experiment group, but not in the control animals. Cluster analysis revealed three distinct clusters of independent features. The features with the lowest variance were Energy, Short Run Emphasis, and Gray Level Non-Uniformity. Conclusions: Texture features may quantitatively detect liver metastases before they become visually detectable by the radiologist. Keywords: Colorectal neoplasms, Computer-assisted image processing, Liver, Magnetic resonance imaging (MRI), Neoplasm micrometastases Key points Colorectal cancer, for example, the entity being the sec- ond highest cause of death in men and women suffering Texture features change systematically in livers with from cancer in the Western world , spreads to the (micro)metastases liver in about 60% of patients and this is often the reason Three clusters of features independently correlated patients ultimately succumb to their disease [2, 3]. Sec- with tumor growth ondary tumors of the liver, therefore, are still a devastat- Texture features may quantitatively detect hepatic ing disease and herald poor prognosis. Fortunately, micrometastases before they become visually interventional as well as surgical techniques for treating detectable liver metastases have made tremendous advances in the last few years [4–6]. However, if a curative approach is chosen, preoperative imaging is essential to correctly Background identify all tumor lesions and avoid leaving behind small The liver is the primary site of distant hematogenous tumor nodules in the future liver remnant. Furthermore, metastases for cancers of the gastrointestinal tract. in postoperative settings, early and correct diagnosis of recurrent tumor lesions is essential for timely treatment * Correspondence: email@example.com decisions such as salvage chemotherapy or repeat sur- Equal contributors gery. Hence, today, in most cancer centers, magnetic Institute of Diagnostic and Interventional Radiology, University Hospital resonance imaging (MRI) of the liver is an integral part Zurich, Raemistrasse 100, 8091 Zurich, Switzerland Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Becker et al. European Radiology Experimental (2018) 2:11 Page 2 of 10 of the workup of patients at risk for liver metastases. Methods Although scan protocols and parameters vary between institutions, T1-weighted and T2-weighted anatomical Animal experiments sequences with high spatial resolution are required . All experiments were carried out in conformity with the Usually, several contrast-enhanced sequences as well as local laws and regulations and had been approved by the diffusion-weighted sequence are included as well . Cantonal veterinary authorities of Zurich before the trial Moreover, the advent of intracellular contrast media start. Male C57BL6 mice aged 8–10 weeks, purchased shows promising results in differentiating metastases from Harlan (Horst, The Netherlands), were used for all from primary liver lesions . experiments. Animals were kept on a 12:12-h day-night Texture analysis is a versatile mathematical technique cycle with water and standard rodent chow provided ad in the field of image analysis established in the seven- libitum. Injections of tumor cells as well as MRI scans ties of the past century [10, 11] and expanded in the were conducted between 8 AM and 12 AM. subsequent decades [12, 13]. In recent years, there has been increasing interest in computing texture features Experimental design from medical images for quantitative analysis called the Mice were injected with MC-38 tumor cells (n =8) or “radiomics” approach . In liver, computed tomog- phosphate-buffered saline (PBS) as controls (sham, n =2). raphy texture-based differentiation between normal tis- The animals underwent MRI before the injection sue, benign tumors, and hepatocellular carcinoma has (baseline) and at days 4, 8, 12, 16, and 20 post injection. been demonstrated to be possible [15, 16]. Hepatic MRI The study duration was set after a pilot series (three ani- texture analysis is able to differentiate healthy from cir- mals, not included in the current analysis) showed defin- rhotic liver  and even quantify the degree of liver fi- itely visible liver tumors on MRI after day 20 post brosis . As texture analysis is not only able to injection. At day 8, two animals of the tumor injection detect morphological lesions but also subtle distortions group were sacrificed to ensure tumor growth by micro- of the tissue architecture, we hypothesized that quanti- scopic examination. At day 20, the remaining animals tative texture-based analysis of MRI (a radiomics ap- were sacrificed and the livers harvested for histologic proach) can identify small niduses of tumor cells earlier examination. The study design is illustrated in Fig. 1b. than qualitative evaluation by the human eye. The purpose of this study was to investigate whether Tumor cell culture any texture features show a correlation with tumor The murine colon cancer cell line MC-38, syngeneic on a growth before the metastasis can be diagnosed in a C57BL6 background, was used for the experiments. Cells human readout based on morphological changes in the were cultured in Dulbecco’s modified eagle medium (Life images. Technologies, Zug, Switzerland) supplemented with 10% Fig. 1 a Scheme (top) and photography (bottom) of the microsurgical intraportal tumor cell or saline (sham) injection. The portal vein being injected is marked with a white arrow. b Experimental study design. MRI was performed at baseline (before injection) and at days 4, 8, 12, 16, and 20 post injection. Two animals from the tumor group were sacrificed for histological examination at day 8 Becker et al. European Radiology Experimental (2018) 2:11 Page 3 of 10 fetal bovine serum and 100 U/mL of penicillin and strepto- All slides were afterwards scanned with a NanoZoomer mycin and incubated at 37 °C and 5% CO . Cell lines were XR Digital slide scanner C12000 (Hamamutsu, Japan) and tested negative for mycoplasma at culture onset (PCR analyzed with the freely available software NDP.view2 Mycoplasma Test Kit; PromoCell, Heidelberg, Germany). (Version 2.6.13, Hamamutsu, Japan). Each slide was separ- Fur tumor cell injections, cells below passage 10 were har- ately scanned for tumor lesions in the whole depicted liver vested by trypsinization, counted with a nucleocounter parenchyma. Area (μm ) and perimeter (μm) of each (Nucleocounter NC-200TM; ChemoMetec A/S, Allerod, tumor lesion were measured with the Freehand Region of Denmark), and prepared in solution of 10 cells/mL PBS. Interest Tool of the NDP.view2 software, as well as the total amount of detected tumor lesions in all slides of each Mouse model and surgical procedures individual animal calculated. An established model of intraportal injection of syngen- eic tumor cells was used for induction of liver tumors as MRI described by Limani et al. . However, in our study, All mice underwent abdominal MRI examinations in a cells were non-selectively injected in all liver lobes. All dedicated small animal 4.7-T scanner (Bruker 4.7-T Phar- animal procedures were undertaken by a surgical maScan 47/16 US, Bruker BioSpin MRI GmbH, Ettlingen, researcher with extensive experience in advanced experi- Germany) under general anesthesia with isoflurane mental microsurgery (MAS). Anesthesia was induced (Attane; Minrad I, Buffalo, NY; 2–3% mixed with pure with isoflurane inhalation (Attane, Minrad I, Buffalo, oxygen). Spin excitation and signal reception were per- NY, USA) 2–3% mixed with pure oxygen; intraoperative formed with a linearly polarized H whole-body mouse analgesia was administered via subcutaneous application coil. The mice were placed in supine position in the scan- of buprenorphine (0.1 mg/kg body weight). Median ner bed and kept warm with a pad circulating a continu- laparotomy of approximately 3-cm length was performed ous supply of warm water during continuous anesthesia. after fixation of the animal with tape on a heating pad. MRI was performed during free breathing with respiratory The liver was mobilized by cutting the falciform liga- control. A T2-weighted rapid acquisition with refocused ment and the membrane between caudate and left lateral echoes sequence was acquired in transverse orientation lobe with microsurgical scissors. After display of the por- with the following parameters: echo time = 19 ms; tal vein, 1 × 10 MC-38 tumor cells, prepared in 100 μL repetition time = 1000 ms; echo-train length = 4; pixel PBS, were injected intraportally with a 29-gauge insulin bandwidth = 310 Hz/pixel; excitations = 2; matrix size syringe (12.7-mm needle length; BD Microfine, Franklin = 192 × 192; field of view = 30 × 30 mm; slice thick- Lakes, NJ, USA) as depicted in Fig. 1a. The needle was ness = 1.5 mm. The images at each time point were then removed and hemostasis achieved by gentle pres- evaluated qualitatively by two independent readers sure with cotton swabs and application of small pieces (ASB, AB) for visibility of metastases. From the visible of Tachosil® (Baxter Inc., Deerfield, IL, USA), if neces- metastases at day 20, in each mouse, one metastasis sary. The abdomen was closed with two layered continu- not yet visible at day 16 was chosen, with easily repro- ous sutures with 5-0 prolene. Mice were allowed to ducible slice position due to anatomical landmarks. recover on a warmed heating pad; food and water were provided 1 h after the operation. Postoperative analgesia Signal-to-noise and contrast-to-noise evaluation with buprenorphine was administered via drinking water The signal-to-noise ratio (SNR) was determined as follows: for three days. Livers were harvested at indicated time pﬃﬃﬃ points (see below) under anesthesia and analgesia as de- SI 2 SNR ¼ scribed above. After re-opening of median laparotomy, noise animals were euthanized by bilateral pneumothorax and trans-section of inferior vena cava and aorta. Organs where SI is the signal intensity in either the liver paren- were harvested quickly and immediately stored in 4% chyma and noise representing the standard deviation in formaldehyde in PBS (% volume/volume). the background (air) measured in the corner of the image outside areas of artefacts. Contrast-to-noise ratio Histological examination (CNR) was defined as: After storage in 4% formaldehyde for 48 h, whole livers were embedded in paraffin blocks in a position resembling SI −SI meta liver transversal slices of the MRI. The whole block was after- CNR ¼ noise wards cut with a cryotome and representative histological slides containing liver and tumor tissue prepared at every millimeter. Slides were colored with hematoxylin-eosin With SI and SI meaning the signal intensity in meta liver (H&E) staining according to standard protocols. the metastases and liver parenchyma, respectively. Becker et al. European Radiology Experimental (2018) 2:11 Page 4 of 10 Texture analysis Statistical analysis Texture analysis was performed with an in-house de- Statistical analysis was performed using the “R” software veloped MATLAB routine (v2016, The MathWorks (v3.3.1., The R Foundation for Statistical Software, Inc., Natick, MA, USA) by the same readers in con- Vienna, Austria). Graphs were generated using “ggplot2” sensus. On a single day-20 slice (acquired at post- . All features were evaluated over the whole time injection day 20), a quadrangular 32 × 32 pixel region course with a linear model/Pearson correlation test. A p of interest (ROI) was placed in the liver, encompass- value < 0.05 was considered statistically significant. The ing a distinct metastasis as illustrated in Fig. 2.The p value was not corrected for multiple comparisons due ROI was manually copied to the same slice at the to the exploratory nature of the analysis. However, the four earlier time points at the same position, with the number of features was reduced with the following three help of anatomical landmarks if the metastasis itself steps: 1 = significantly changing features in the sham was not visible. From the two control animals (sham), group were excluded from the final set; 2 = features two and three slices were analyzed in order to yield were examined for redundancy by co-correlation testing five data points and reasonable confidence intervals. (Pearson) and hierarchical clustering to determine Before texture analysis, ROI contents were normalized groups of independently changing features; and 3 = from between the mean and three standard deviations to each cluster, the feature with the smallest variance was minimize intra- and inter-scanner fluctuations in tex- selected as the most representative one. ture analysis . Thirty-two texture features were computed: four Results first-order and 28 higher-order features analogous to Study procedures those described by Becker et al.  and Vallières et The intraportal tumor cell and sham injections were al. , as summarized in Table 1. The first order performed successfully and without any complications. features were computed directly from the histogram MRI scans before injection (baseline) and at days 4, 8, of the original image, whereas the higher order 12, 16, and 20 after injection of MC38 tumor cells were features were obtained from the gray-level co- completed successfully. Presence of tumor cells in the occurrence matrix (GLCM), the gray-level run-length liver parenchyma was confirmed histologically after matrix (GLRLM), or the gray-level size zone matrix eight days in two mice, which were sacrificed for this (GLSZM). Albeit some of these features have “intui- purpose (Fig. 3). At post-injection day 20, T2-weighted tive names” (“intuitive” in this context meaning easily images showed well visible hyperintense liver tumors in distinguishable by the human observer), none resem- all six remaining mice (Fig. 2,bottomright). ble or describe any intuitive patterns . The math- ematical definition of the respective features can be Morphological evaluation found in the works by Haralick et al. for the GLCM On MRI, the mice exhibited a median of four metas- , Mary M. Galloway for the GLRLM , and tases on day 20 (range = 3–11). In each animal, there Thibault et al. for the GLSZM . was at least one metastasis near an anatomic Fig. 2 Sample slices of an animal of the experiment group. The metastases are well delineated after 20 days but not definitely visible beforehand. The three vessel branches near the ROI (arrowhead) serve as an anatomical landmark to analyze the same volume of liver tissue in the images before day 20, when the metastasis is not visible yet Becker et al. European Radiology Experimental (2018) 2:11 Page 5 of 10 Table 1 Texture features used in the present study Primary Higher order Histogram Gray-level co-occurrence matrix (GLCM) Gray-level run-length matrix (GLRLM) Gray-level size-zone matrix (GLSZM) Variance Contrast Short run emphasis (SRE) Small zone emphasis (SZE) Skewness Correlation Long run emphasis (LRE) Large zone emphasis (LZE) Kurtosis Energy Gray-level non-uniformity (GLN) Gray-level non-uniformity (GLN) Entropy Homogeneity Run length non-uniformity (RLN) Zone-size non-uniformity (ZSN) Run percentage (RP) Zone percentage (ZP) Low gray-level run emphasis (LGRE) Low gray-level zone emphasis (LGZE) High gray-level run emphasis (HGRE) High gray-level zone emphasis (HGZE) Short run low gray-level emphasis (SRLGE) Small zone low gray-level emphasis (SZLGE) Short run high gray-level emphasis (SRHGE) Small zone high gray-level emphasis (SZHGE) Long run low gray-level emphasis (LRLGE) Large zone low gray-level emphasis (LZLGE) Long run high gray-level emphasis (LRHGE) Large zone high gray-level emphasis (LZHGE) Gray level variance (GLV) Zone size variance (ZSV) Fig. 3 Representative histological images of mouse livers (H&E staining). RML right median lobe, GB gall bladder, LML left median lobe, LLL left lateral lobe, CL caudate lobe, RL right inferior and superior lobe. Orange squares mark the area of 40× magnification for the respective images below. a Overview of mouse liver with control PBS injection, harvested at day 20. Overview in 1.25× magnification shows complete transversal section of the liver covering RML, GB, LML, LLL, and the bifid CL. 10× magnification and 40× magnification show intact liver parenchyma without any signs of tumor invasion. b Overview of a mouse liver harvested on day 8 after non-selective intraportal injection of syngeneic MC38 tumor cells. While no tumor can be detected macroscopically and in the overview of the specimen, 10× and 40× magnification reveal small nests of intraparenchymal and paravascular tumor cells, accompanied by infiltrating leukocytes. c Overview of mouse liver harvested on day 20 after non-selective intraportal injection of syngeneic MC38 tumor cells. Multiple tumor nodules can be appreciated already at a macroscopic level in all liver lobes Becker et al. European Radiology Experimental (2018) 2:11 Page 6 of 10 landmark which was reliably depicted on all days and Table 2 Correlating features with Pearson correlation coefficients (R) and p values thus suitable for texture analysis. On histology on day 8, the median circumference was 0.173 mm (interquartile Feature R p value range [IQR] = 0.137–0.205 mm) corresponding to an area Variance 0.204 0.194 2 2 of 0.001812 mm (IQR = 0.001243–0.002513 mm ); on Skewness 0.198 0.210 day 20, the circumference had grown to 3.57 mm (IQR = Kurtosis 0.411 0.007 2.19–9.27 mm) which corresponds to an area of 0.61 mm Contrast (GLCM) − 0.361 0.019 (IQR = 0.24–5.10 mm ). Correlation (GLCM) 0.393 0.010 Energy (GLCM) 0.392 0.010 Signal-to-noise and contrast-to-noise ratios SNRs were (mean ± standard deviation) 28.98 ± 6.52 (day Homogeneity 0.432 0.004 0), 23.71 ± 9.90 (day 4), 28.38 ± 6.98 (day 8), 26.44 ± 6.23 Entropy (GLCM) − 0.014 0.930 (day 12), 26.72 ± 7.32 (day 16), and 28.08 ± 8.15 (day 20). SRE (GLRLM) − 0.394 0.010 CNR of metastases on day 20 was 6.88 ± 4.63. LRE (GLRLM) 0.410 0.007 GLN (GLRLM) 0.419 0.006 Texture analysis RLN (GLRLM) − 0.398 0.009 Texture features were computed successfully for all time points and animals. Linear fitting revealed significant RP (GLRLM) − 0.406 0.008 correlation in 18 features in the experiment group, as LGRE (GLRLM) − 0.103 0.516 follows (full names in Table 1): HGRE (GLRLM) − 0.129 0.417 SRLGE (GLRLM) − 0.110 0.486 – First order: Kurtosis SRHGE (GLRLM) − 0.286 0.067 – GLCM: Contrast, Correlation, Energy, Homogeneity LRLGE (GLRLM) − 0.089 0.575 – GLRLM: SRE, LRE, GLN, RLN, RP, LRHGE – GLSZM: SZE, LZE, GLN, ZSN, ZP, SZHGE, LZHGE LRHGE (GLRLM) 0.407 0.008 SZE (GLSZM) − 0.447 0.003 Correlation coefficients and p values are summarized LZE (GLSZM) 0.386 0.011 in Table 2. GLN (GLSZM) 0.386 0.011 A selected set of those features is shown in Fig. 4. Five ZSN (GLSZM) − 0.445 0.003 features correlated significantly in the sham group: ZP (GLSZM) − 0.428 0.005 – GLRLM: LGRE, SRLGE LGZE (GLSZM) − 0.092 0.561 – GLSZM: LGZE, GLV, ZSV HGZE (GLSZM) − 0.182 0.248 SZLGE (GLSZM) − 0.117 0.462 However, none of them were also significantly corre- SZHGE (GLSZM) − 0.345 0.025 lated in the experiment group. LZLGE (GLSZM) − 0.051 0.750 Hierarchical clustering revealed three distinct, inde- LZHGE (GLSZM) 0.383 0.012 pendent clusters of features as depicted in Fig. 5.The most representative features, i.e. the ones with the smal- GLV (GLSZM) − 0.269 0.085 lest variance were Energy, SRE (GLRLM), and GLN ZSV (GLSZM) − 0.086 0.589 (GLSZM). Discussion Recently published studies have demonstrated that In the present study, we examined whether changes of texture analysis can distinguish or classify benign and texture features may herald metastases in liver MRI be- malignant lesions in various organs and tumors, for ex- fore they can be appreciated visually. We found three ample in glioma/glioblastoma [25, 26], breast , lung independent features, one derived from each of the , stomach , prostate , or liver lesions [15, 31]. gray-level matrices, which exhibit a linear correlation Another recent focus of texture analysis has been the as- before the metastasis is visible to the naked eye, and sessment of therapy response, e.g. in advanced ovarian several co-dependent features. Thereby, we showed that and primary peritoneal cancer , or the prediction of texture analysis is able to detect subtle changes of par- lymph node metastasis from a radiomics analysis of the enchymal changes before a morphological lesion is vis- primary tumor . However, to the best of our know- ible, which may significantly enhance tumor detection ledge, no study has so far investigated the feasibility to rates in liver imaging. detect cancerous lesions directly in the target tissue Becker et al. European Radiology Experimental (2018) 2:11 Page 7 of 10 Fig. 4 Set of features which change significantly after the injection of tumor cells at day 0, but not in the control group after sham injection of PBS before they appear visible to the human reader. In our adjunctive value in these cases due to the high opinion, this application logically follows from the background glycolytic activity of the liver . common denominator of the abovementioned studies: Texture analysis may thus be a new objective method quantifying underlying histological changes in tissue to detect these lesions and improve post-surgery out- below the resolution of the given modality or comes and disease-free survival interval. On the basis of protocol. the current data, it is not possible to determine whether Leonard et al. aswellasAdam etal. showed the textural changes are a result of the metastatic cells that an increased number of patients with liver metastases themselves or rather a reflection of reactive changes in undergo potentially curative hepatic resection because of the surrounding liver parenchyma. Interestingly, features recent progress in neoadjuvant chemotherapy. Still, opti- derived from all three gray-level matrices appear to be mal surgical planning depends on exact knowledge of the influenced by the metastatic growth, which could be an number and location of all liver lesions. Recently indication for the destruction of liver acini (alteration of published studies showed high diagnostic accuracy for the co-occurrence and size-zones) or the tumor neovascular- detection of liver metastases in modern imaging modal- ization (run-lengths of vessels). Further research in this ities such as MRI and PET/MRI . However, about area may be desirable as understanding the exact mech- two-thirds of the patients who have undergone liver resec- anism may aid for example in development of better tion for colorectal metastases suffer from recurrence MRI sequences suitable for texture analysis. within 18 months . One reason is probably the fact Our study has several limitations that need to be ac- that small liver metastases below or close to the resolution knowledged. First, although the images were prospect- limit of the current imaging modalities on pre-surgery im- ively acquired, ROI definition had to be performed aging are missed and, therefore, not taken into account. retrospectively after a suitable lesion was identified at F-fluorodeoxyglucose (FDG) PET/CT is of little the study end. Furthermore, we have only evaluated Becker et al. European Radiology Experimental (2018) 2:11 Page 8 of 10 Fig. 5 Correlation matrix showing the redundancy of many features. However, some clusters of independent features can be identified. The Pearson correlation coefficient R is color-coded according to the scale on the right single slices. Because we aimed for a maximum in-plane be more experiments in animals, but rather a longitu- resolution, the sequence was not acquired with isotropic dinal study directly in human patients, e.g. a cohort at voxel size and respective slice gaps. Performing three- risk for hepatic (colorectal cancer) metastases. Until fur- dimensional texture analysis would have either required ther validation in human studies, the implications of this interpolation (which has been shown to confound the work for patient care remain unclear. analysis ) or a lower resolution. Thus, we believe that In conclusion, we found in our small pilot study that three-dimensional analysis would not have added value texture analysis of MRI data may have the potential to to our results or altered our conclusion. Second, we only detect liver metastases at a sub-resolution level, before computed a limited set of features. We chose to do so, they become visible to the human eye. instead of analyzing a larger set of multiple hundred or Abbreviations thousand (compound) features, because the selected fea- GLCM: Gray-level co-occurrence matrix; GLRLM: Gray-level run-length matrix; tures have repeatedly been found useful in the analysis GLSZM: Gray-level size-zone matrix; MRI: Magnetic resonance imaging; of medical images [15, 28, 30, 40] and robust against var- PBS: Phosphate-buffered saline; ROI: Region of interest iations between scanners and protocol parameters , Acknowledgements especially after normalization . Moreover, our small The authors thank the employees of the animal laboratory of the University sample size did not allow us to use multiparametric/hy- hospital of Zurich for their important contributions to this work. brid imaging or machine learning algorithms to assess the usefulness of such a large number of features, which Funding ASB was partly funded by the clinical research priority program molecular is the third main limitation. However, adhering to the imaging network Zurich (CRPP MINZ). None of the remaining authors have 3R-principle (“Replace-Reduce-Refine”), the small any relevant funding to declare. number of animals was a deliberate effort to keep the suffering of animals as low as possible. Hence, we think Availability of data and materials that the next step after this small pilot study should not All relevant data are presented in the article. Becker et al. European Radiology Experimental (2018) 2:11 Page 9 of 10 Guarantor of the study 15. Mougiakakou SG, Valavanis IK, Nikita A et al (2007) Differential diagnosis of Andreas Boss, M.D., Ph.D. CT focal liver lesions using texture features, feature selection and ensemble driven classifiers. Artif Intell Med 41:25–37 Authors’ contributions 16. Bharathi VS, Ganesan L (2008) Orthogonal moments based texture analysis ASB conceived the idea. ASB, MAS, AB planned the study. MAS, PAC obtained of CT liver images. Pattern Recogn Lett 29:1868–1872 approval for the animal experiments. MAS performed the microsurgical 17. Jirak D, Dezortová M, Taimr P et al (2002) Texture analysis of human liver. procedures. ASB, MCW, AB performed MRI acquisition. ASB, MCW, MWW J Magn Reson Imaging 15:68–74 performed image post-processing (texture analysis). ASB, AB performed the 18. Barry B, Buch K, Soto JA et al (2014) Quantifying liver fibrosis through the statistical analysis. ASB, MAS performed an in-depth literature research and application of texture analysis to diffusion weighted imaging. Magn Reson wrote the first draft of the manuscript. All authors were involved in editing the Imaging 32:84–90 manuscript and approved the final version. 19. Limani P, Borgeaud N, Linecker M et al (2016) Selective portal vein injection for the design of syngeneic models of liver malignancy. Am J Physiol Ethics approval and consent to participate Gastrointest Liver Physiol 310:G682–G688 This study was approved by the Cantonal veterinary authorities of Zurich. 20. Collewet G, Strzelecki M, Mariette F (2004) Influence of MRI acquisition protocols and image intensity normalization methods on texture Consent for publication classification. Magn Reson Imaging 22:81–91 Not applicable 21. Becker AS, Wagner MW, Wurnig MC et al (2017) Diffusion-weighted imaging of the abdomen: Impact of b-values on texture analysis Competing interests features. NMR Biomed 30:1 The authors declare that they have no competing interests. 22. Vallières M, Freeman C, Skamene S et al (2015) A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metastases in soft-tissue sarcomas of the extremities. Phys Med Biol 60:5471 Publisher’sNote 23. Haralick RM, Shanmugam K, Dinstein I (1973) Textural features for image Springer Nature remains neutral with regard to jurisdictional claims in classification. IEEE T Syst Man Cyb SMC-3:610–621 published maps and institutional affiliations. 24. Wickham H (2009) Ggplot2. Elegant graphics for data analysis. Springer- Verlag, New York Author details 25. Skogen K, Schulz A, Dormagen JB et al (2016) Diagnostic performance Institute of Diagnostic and Interventional Radiology, University Hospital of texture analysis on MRI in grading cerebral gliomas. Eur J Radiol 85: Zurich, Raemistrasse 100, 8091 Zurich, Switzerland. Division of 824–829 Transplantation and Visceral Surgery, University Hospital Zurich, Raemistrasse 26. Hu LS, Ning S, Eschbacher JM et al (2015) Multi-parametric MRI and texture 100, 8091 Zurich, Switzerland. analysis to visualize spatial histologic heterogeneity and tumor extent in glioblastoma. PLoS One 10:e0141506 Received: 6 March 2018 Accepted: 11 April 2018 27. Waugh S, Purdie C, Jordan L et al (2016) Magnetic resonance imaging texture analysis classification of primary breast cancer. Eur Radiol 26: 322–330 References 28. Dennie C, Thornhill R, Sethi-Virmani V et al (2016) Role of quantitative 1. Haggar FA, Boushey RP (2009) Colorectal cancer epidemiology: Incidence, computed tomography texture analysis in the differentiation of primary mortality, survival, and risk factors. Clin Colon Rectal Surg 22:191–197 lung cancer and granulomatous nodules. Quant Imaging Med Surg 6:6 2. Weiss L, Grundmann E, Torhorst J et al (1986) Haematogenous metastastic 29. Ba-Ssalamah A, Muin D, Schernthaner R et al (2013) Texture-based patterns in colonic carcinoma: An analysis of 1541 necropsies. J Pathol 150: classification of different gastric tumors at contrast-enhanced CT. Eur J 195–203 Radiol 82:e537–e543 3. Sadahiro S, Suzuki T, Ishikawa K et al (2002) Recurrence patterns after 30. Wibmer A, Hricak H, Gondo T et al (2015) Haralick texture analysis of prostate MRI: curative resection of colorectal cancer in patients followed for a minimum Utility for differentiating non-cancerous prostate from prostate cancer and of ten years. Hepatogastroenterology 50:1362–1366 differentiating prostate cancers with different Gleason scores. Eur Radiol 25:2840–2850 4. Clavien P-A, Petrowsky H, DeOliveira ML et al (2007) Strategies for safer liver 31. Mayerhoefer ME, Schima W, Trattnig S et al (2010) Texture-based surgery and partial liver transplantation. New Engl J Med 356:1545–1559 classification of focal liver lesions on MRI at 3.0 Tesla: A feasibility study in 5. Schnitzbauer AA, Lang SA, Goessmann H et al (2012) Right portal vein cysts and hemangiomas. J Magn Reson Imaging 32:352–359 ligation combined with in situ splitting induces rapid left lateral liver lobe 32. Kyriazi S, Collins DJ, Messiou C et al (2011) Metastatic ovarian and primary hypertrophy enabling 2-staged extended right hepatic resection in small- peritoneal cancer: Assessing chemotherapy response with diffusion- for-size settings. Ann Surg 255:405–414 weighted MR imaging—value of histogram analysis of apparent diffusion 6. de Santibañes E, Clavien P-A (2012) Playing play-doh to prevent coefficients. Radiology 261:182–192 postoperative liver failure: The “alpps” approach. Ann Surg 255:415–417 33. Huang Y, Liang C, He L et al (2016) Development and validation of a 7. The American College of Radiology (2014) Liver Imaging Reporting and radiomics nomogram for preoperative prediction of lymph node metastasis Data System. Version 2014. https://nrdr.acr.org/lirads/. Accessed 31 Mar 2018 in colorectal cancer. J Clin Oncol 34:2157–2164 8. Donati OF, Fischer MA, Chuck N et al (2013) Accuracy and confidence of 34. Leonard GD, Brenner B, Kemeny NE (2005) Neoadjuvant chemotherapy gd-eob-dtpa enhanced MRI and diffusion-weighted imaging alone and in before liver resection for patients with unresectable liver metastases from combination for the diagnosis of liver metastases. Eur J Radiol 82:822–828 colorectal carcinoma. J Clin Oncol 23:2038–2048 9. Goshima S, Kanematsu M, Watanabe H et al (2010) Hepatic hemangioma 35. Adam R, Aloia T, Lévi F et al (2007) Hepatic resection after rescue cetuximab and metastasis: Differentiation with gadoxetate disodium–Enhanced 3-T treatment for colorectal liver metastases previously refractory to MRI. AJR Am J Roentgenol 195:941–946 conventional systemic therapy. J Clin Oncol 25:4593–4602 10. Galloway MM (1975) Texture analysis using gray level run lengths. Comput 36. Lee DH, Lee JM, Hur BY et al (2016) Colorectal cancer liver metastases: Vision Graph 4:172–179 Diagnostic performance and prognostic value of PET/MR imaging. 11. Tamura H, Mori S, Yamawaki T (1978) Textural features corresponding to Radiology 280:782–792 visual perception. IEEE Trans Syst Man Cybern Syst 8:460–473 37. Penna C, Nordlinger B (2002) Colorectal metastasis (liver and lung). Surg 12. Dasarathy BV, Holder EB (1991) Image characterizations based on joint gray Clin North Am 82:1075–1090 level—run length distributions. Pattern Recogn Lett 12:497–502 13. Thibault G, Fertil B, Navarro C et al (2009) Texture indexes and gray level 38. Yi CA, Shin KM, Lee KS et al (2008) Non–small cell lung cancer staging: size zone matrix application to cell nuclei classification. Pattern Recognition efficacy comparison of integrated PET/CT versus 3.0-T whole-body MR Inf Process 27:1357002 imaging. Radiology 248:632–642 14. Lambin P, Rios-Velazquez E, Leijenaar R et al (2012) Radiomics: Extracting 39. Mayerhoefer ME, Szomolanyi P, Jirak D et al (2009) Effects of magnetic more information from medical images using advanced feature analysis. Eur resonance image interpolation on the results of texture-based pattern J Cancer 48:441–446 classification: A phantom study. Investig Radiol 44:405–411 Becker et al. European Radiology Experimental (2018) 2:11 Page 10 of 10 40. MacKay JW, Murray PJ, Low SB et al (2016) Quantitative analysis of tibial subchondral bone: Texture analysis outperforms conventional trabecular microarchitecture analysis. J Magn Reson Imaging 43:1159–1170 41. Mayerhoefer ME, Szomolanyi P, Jirak D et al (2009) Effects of MRI acquisition parameter variations and protocol heterogeneity on the results of texture analysis and pattern discrimination: An application-oriented study. Med Phys 36:1236–1243 42. Flecknell P (2002) Replacement, reduction and refinement. ALTEX 19:73–78
European Radiology Experimental – Springer Journals
Published: May 28, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera