ELSEVIER
Contents lists available at ScienceDirect
Clinical Radiology
-
a
clinical ŘADÍOLOGY
MRCR
Radiomic mapping model for prediction of Ki-67 expression in adrenocortical carcinoma
A.A. Ahmed ª, M.M. Elmohr b, D. Fuentes b, M.A. Habra , S.B. Fisherd, N.D. Perrierª, M. Zhang e, K.M. Elsayes ª,*
a Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
b Department Imaging Physics, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
” Department Endocrine Neoplasia and Hormonal Disorders, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
d Department Surgical Oncology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
e Department Pathology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA
ARTICLE INFORMATION
Article history: Received 18 September 2019 Accepted 23 January 2020
AIM: To determine the value of contrast-enhanced computed tomography (CT)-derived radiomic features in the preoperative prediction of Ki-67 expression in adrenocortical carci- noma (ACC) and to detect significant associations between radiomic features and Ki-67 expression in ACC.
MATERIALS AND METHODS: For this retrospective analysis, patients with histopathologically proven ACC were reviewed. Radiomic features were extracted for all patients from the pre- operative contrast-enhanced abdominal CT images. Statistical analysis identified the radiomic features predicting the Ki-67 index in ACC and analysed the correlation with the Ki-67 index.
RESULTS: Fifty-three cases of ACC that met eligibility criteria were identified and analysed. Of the radiomic features analysed, 10 showed statistically significant differences between the high and low Ki-67 expression subgroups. Multivariate linear regression analysis yielded a pre- dictive model showing a significant association between radiomic signature and Ki-67 expression status in ACC (R2=0.67, adjusted R2=0.462, p=0.002). Further analysis of the in- dependent predictors showed statistically significant correlation between Ki-67 expression and shape flatness, elongation, and grey-level long run emphasis (p=0.002, 0.01, and 0.04, respectively). The area under the curve for identification of high Ki-67 expression status was 0.78 for shape flatness and 0.7 for shape elongation.
CONCLUSION: Radiomic features derived from preoperative contrast-enhanced CT images show encouraging results in the prediction of the Ki-67 index in patients with ACC. Morphological features, such as shape flatness and elongation, were superior to other radiomic features in the detection of high Ki-67 expression.
@ 2020 Published by Elsevier Ltd on behalf of The Royal College of Radiologists.
* Guarantor and correspondent: K. M. Elsayes, Department of Radiology, The University of Texas MD Anderson Cancer Center, 1400 Pressler Street, Houston, TX 77030, USA.Tel .: +1 713-745-3025; fax: +1 713-794-4379.
E-mail address: kmelsayes@mdanderson.org (K.M. Elsayes).
Introduction
Adrenocortical carcinoma (ACC) is a rare, highly aggres- sive tumour of the adrenal cortex with a reported annual incidence of one case per million population. This tumour is highly fatal and has a very poor prognosis, with 5-year overall survival rates ranging from 14% to 44%.1 The only curative treatment for ACC is surgical resection, which is considered the mainstay treatment for stage I-III disease.2 Even for surgically resectable disease, there is a high risk for local recurrence, and the outcome is highly variable.3,4
Ki-67 is a protein exclusively expressed by proliferating cells and is, therefore, used routinely as a marker of cellular proliferation for multiple tumours, including ACC.5,6 Given its prognostic significance, the Ki-67 index is considered one of the most important established prognostic markers for local recurrence of ACC7; however, the Ki-67 index can be quantified only through histopathological examination of resected tumour tissue8; biopsy is not recommended in adrenal masses because of the fallacies in differentiating benign from malignant lesions. Furthermore, core biopsy samples are not considered a true indicator of tumour Ki-67 expression because of the heterogeneity of Ki-67 expression in different regions of the tumour; a low Ki-67 value, for example, if detected in a core biopsy sample, might be a misleading result.9
Medical imaging, including computed tomography (CT), magnetic resonance imaging (MRI), and positron-emission tomography (PET), is used for almost all oncology patients during their diagnosis, treatment, and follow-up moni- toring. Although these images play a crucial role in the diagnosis and staging of different tumours, there are limi- tations in the assessment of the degree of tumour hetero- geneity.10,11 Extraction of radiomic features and texture analysis is the process of converting digital medical images into quantitative data through a set of mathematical cal- culations, which simply quantify the degree of tumour heterogeneity; through repeated use of this method in all oncological settings, signatures of radiomic features that accurately reflect the heterogeneous biological behaviour of tumours have been developed.10,12
CT texture analysis (CTTA) is a novel, emerging technique that relies on quantifying the heterogeneity of a tumour to predict its progression, prognosis, and response to treat- ment.13,14 Multiple studies were conducted using CTTA to predict disease outcome in different oncological settings, such as lung, and head and neck cancers12; however, there are no studies on the correlation between CTTA results and Ki-67 expression in ACC. Therefore, the aim of the present study was to define the utility of CTTA to non-invasively predict the Ki-67 index in ACC.
Materials and methods
Study design and cohort
The present study was approved by the Institutional Review Board, and the requirement for informed consent
was waived. Clinical and imaging data were collected retrospectively from the electronic medical records to include patients fulfilling the following criteria: (1) diag- nosis of ACC, (2) underwent surgical resection of the tumour, (3) the Ki-67 index was determined as part of the histopathological evaluation of the resected tissue, (4) im- aging data (pre-resection contrast-enhanced CT of the abdomen) were available. Patients whose Ki-67 was quan- tified in biopsy tissue, rather than resected whole tumour, were excluded from the study. This exclusion was based on previous studies concluding that Ki-67 quantification should be based on tissue samples collected from the whole tumour.9,15
A total of 53 patients with ACC with a mean age at diagnosis of 53 years, including 31 women and 22 men were enrolled in the study (Fig 1); all cases were diagnosed be- tween 2006 and 2018. Tumour characteristics and patient demographics are summarised in Table 1.
CT image acquisition
Preoperative CT was obtained in the venous phase (60-80 seconds after intravenous contrast medium injec- tion). Some of the scans included in the present study were performed in outside facilities. The scans performed at MD Anderson Cancer Center were acquired on 64-multidetector CT Light-Speed scanners (GE Healthcare, Waukesha, WI, USA) with a section thickness of 2.5 mm and an injection rate of 3-5 ml/s.
Radiomic feature extraction
The scans were exported from the picture archiving and communication system (PACS) in the digital imaging in communications and medicine (DICOM) format, and then converted to the Neuroimaging Informatics Technology Initiative (NIFTI) format.16 The adrenal gland, including the malignant tissue, was manually segmented using the AMIRA software package.17 Images were exported to the PyRadiomics platform version 2.1.1 for image analysis and for extraction of radiomic features.18 For each lesion, 106 radiomic features were computed. The derived features can be categorised into seven classes: first-order statistics (18 features), shape-based features (14 features), grey-level co- occurrence matrix (GLCM: 23 features), grey-level run length matrix (GRLRM; 16 features), neighbouring grey tone difference matrix (NGTDM; five features), grey-level dif- ference matrix (GLDM; 14 features), and grey-level size zone matrix (GLSZM; 16 features). Details of the definition and the extraction of the radiomic features are published.10,11,18
Ki-67 labelling index
Sections of all resected tissues had been immunostained for the Ki-67 proliferation marker using the MIB-1 antibody clone. Areas with brown nuclear staining were considered positive, while unstained areas were considered negative. The three to four areas with highest positive staining were selected for examination. The Ki-67 index was reported as
Please cite this article as: Ahmed AA et al., Radiomic mapping model for prediction of Ki-67 expression in adrenocortical carcinoma, Clinical Radiology, https://doi.org/10.1016/j.crad.2020.01.012
Inclusion criteria:
From Jan. 2006 to Dec. 2018 345 patients with pathologically diagnosed adrenocortical carcinoma (ACC) who underwent preoperative contrast-enhanced CT
Exclusion criteria:
Technically inadequate CT studies (N=142)
Unavailable Ki-67 index (N=130) Neo-adjuvant chemotherapy (N=20)
Study group:
53 treatment-naïve patients were included for the prediction of the Ki-67 index
the percentage of the positively stained nuclei in each area. High Ki-67 expression was defined as Ki-67 index >10% and low expression was defined as Ki-67 index <10%.
Statistical analysis
Statistical analyses were carried out using IBM SPSS software version 24.19 Descriptive statistics were calculated to summarise clinical and histopathological data. Clinical data included patient age, sex, race, and presenting clinical symptoms. Age was calculated at the time of the diagnosis. Pathological features included the tumour size, laterality, and disease stage. Disease stage was determined according to the American Joint Committee on Cancer/Union for In- ternational Cancer Control (AJCC/UICC) cancer staging
| Characteristic | Value (n=53) |
|---|---|
| Mean age at diagnosis | 53 years (SD, 13.47 years) |
| Mean tumour size | 11.5 cm (SD, 6.48 cm) |
| Tumour side | |
| Right | 24/53 (45.3%) |
| Left | 29/53 (54.7%) |
| Sex | M = 22, F = 31 |
| Clinical presentation | |
| Incidentaloma | 11/53 (20.8%) |
| Hormonal hypersecretion | 22/53 (41.5%) |
| Compressive symptoms | 20/53 (37.7%) |
| TNM staging | |
| Stage I | 4/53 (7.5%) |
| Stage II | 19/53 (35.8%) |
| Stage III | 23/53 (43.4%) |
| Stage IV | 7/53 (13.2%) |
| Ki-67 index | |
| Low (≤10%) | 13/53 (24.5%) |
| High (>10%) | 40/53 (75.5%) |
F, female; M, male; SD, standard deviation; TNM, tumour, node, metastasis.
manual 8th edition.2º Frequencies and percentages were used for reporting categorical variables, while means and standard deviations were used for continuous ones. The Mann-Whitney test was used for comparing the contin- uous variables between the low and high Ki-67 expression subgroups.
The 106 extracted radiomic features from all patients were entered in a multivariate linear regression model. Ki- 67 value represented the dependent variable, and radiomic features represented the independent variables or the possible predictors. The forward stepwise method was used for feature selection. The p-value thresholds for variables entering and leaving the model were 0.05 and 0.1, respec- tively. Using the high and low levels of Ki-67 already defined receiving operating characteristic (ROC) were con- structed for statistically significant predictors in the multi- variate model, and the area under curve (AUC), sensitivity, specificity, and positive and negative likelihood ratios were calculated. Spearman’s correlation coefficient was used to determine the correlation between the Ki-67 index and each of the radiomic features that were significantly different in the multivariate regression model. A p-value of <0.05 was considered statistically significant.
Results
Among the 53 cases enrolled in the study, 24 patients (45.3%) had right-sided adrenal lesions and 29 (54.7%) had left-sided adrenal lesions. Thirteen of the tumours (24.5%) showed low Ki-67 expression (≤10 %) and 40 (75.5%) showed high Ki-67 expression (>10%) on immunohisto- chemical analysis.
After manual segmentation of the adrenal glands on the CT images, radiomic features of the tumours were analysed. Of the 106 radiomic features extracted, 10 showed a
Please cite this article as: Ahmed AA et al., Radiomic mapping model for prediction of Ki-67 expression in adrenocortical carcinoma, Clinical Radiology, https://doi.org/10.1016/j.crad.2020.01.012
significant intergroup difference between the high and low Ki-67 tumours using Mann-Whitney analysis (Table 2). The overall multivariate linear regression model showed a sta- tistically significant association (R2=0.67, adjusted R2=0.462, r=0.824, and p=0.002) between the radiomic signature derived from contrast-enhanced CT-derived fea- tures and Ki-67 expression in the ACCs. The significantly independent predictors in the present model (Table 3) were shape elongation (p=0.021), shape flatness (p=0.04), GLCM cluster shade (p=0.006), GLRLM long run emphasis (p=0.011), GLSZM (p=0.014), interquartile range (first or- der; p=0.034), and NGTDM (p=0.031). Spearman’s corre- lation analysis showed significant correlations between the Ki-67 index and independent predictors in the model only for shape elongation (r = 0.352, p=0.01), shape flatness (r =0.414, p=0.002), and GLRLM long run emphasis (r =- 0.274, p=0.04).
The AUCs for detecting high Ki-67 expression were 0.7 for shape elongation and 0.78 for shape flatness (Fig 2). The rest of the features showed AUCs <0.5. The cut-off value, sensitivity, specificity, and positive likelihood and negative likelihood ratios were 0.668, 75% 69.2%, 2.4, and 0.36, respectively, for shape elongation and 0.4966, 80%, 69.2%, 2.5, and 0.28, respectively, for shape flatness.
Discussion
In the current study, a radiomic signature was built using pretreatment contrast-enhanced abdominal CT for the prediction of the Ki-67 index in patients with ACC and detected the significant associations between radiomic features and Ki-67 expression in ACC. The present results suggest that contrast-enhanced CT-derived radiomic fea- tures have utility in the prediction of Ki-67 expression in ACCs. Ki-67 is a proliferation marker with an established prognostic role in a myriad of tumours, including ACC.6,7,21,22 The kinetics of intratumour heterogeneity are likely related to the degree of cellular proliferation of ma- lignant cells, and thus Ki-67 expression.23 A study in a large Italian/German cohort concluded that Ki-67 was the single most important indicator of recurrence in ACC after com- plete resection.7
| Radiomic feature (class) | Standardised regression coefficientª | p-Value |
|---|---|---|
| Cluster shade (GLCM) | 12.859 | 0.006 |
| Long run emphasis (GLRLM) | 8.946 | 0.011 |
| Grey level non-uniformity normalised (GLSZM) | -2.083 | 0.014 |
| Elongation (shape) | 1.112 | 0.021 |
| Contrast (NGTDM) | 4.312 | 0.031 |
| Interquartile range (first order) | -0.75 | 0.034 |
| Flatness (shape) | -1.019 | 0.040 |
A p value < 0.05 was considered statistically significant GLCM, grey-level co-occurrence matrix; GLRLM, grey-level run length ma- trix; GLSZM, grey-level size-zone matrix; NGTDM, neighbouring grey-tone difference matrix.
a Standardised regression coefficient measures the correlation between the independent and dependent variables.
Another study showed that the Ki-67 index was an important prognostic marker for overall survival in patients with stage IV ACC.24 The diagnostic and prognostic value of Ki-67 is well established in various malignancies.25 Owing to its important prognostic value, it is now recommended that the Ki-67 index be included in the histopathological analysis of all resected ACC specimens.9 There is consensus on using a Ki-67 index >10% as a marker of high risk of recurrence, which justifies the use of adjuvant chemo- therapy in this group of patients.9,26 The use of adjuvant chemotherapy, namely mitotane, has been found to improve disease progression and survival; however, given the serious side-effect profile of mitotane, including adrenal crisis, it should be reserved for patients with high risk of recurrence, i.e. Ki-67 index >10%.9
Multiple studies have been published regarding pre- dicting Ki-67 status in different tumours using a radiomic signature derived from either CT or MRI studies.23,27-31 Because little was known about the utility of using radio- mic features derived from contrast-enhanced CT images to predict Ki-67 expression in ACC, the utility of a wide variety of radiomic features was explored, including first-order statistics, shape-based features, and grey-level parame- ters, a total of 106 features for each lesion. The resulting
| Radiomic feature (class) | Low Ki-67 tumours | High Ki-67 tumours | p-Value |
|---|---|---|---|
| Maximum 3D diameter (shape) | 35.92 | 24.10 | 0.016 |
| Major axis length (shape) | 36.85 | 23.80 | 0.008 |
| Elongation (shape) | 19 | 29.6 | 0.032 |
| Flatness (shape) | 15.62 | 30.7 | 0.002 |
| Maximum 2D diameter column (shape) | 35.54 | 24.23 | 0.022 |
| Maximum 2D diameter row (shape) | 36.38 | 23.95 | 0.012 |
| Inverse difference moment normalised (GLCM) | 35.85 | 24.13 | 0.017 |
| Inverse difference normalised (GLCM) | 35.54 | 24.23 | 0.022 |
| Maximum (first order) | 38.92 | 25.73 | 0.001 |
| Strength (NGTDM) | 35.77 | 24.1 | 0.018 |
A p-value <0.05 was considered statistically significant.
GLCM, grey-level co-occurrence matrix; NGTDM, neighbouring grey-tone difference matrix.
Please cite this article as: Ahmed AA et al., Radiomic mapping model for prediction of Ki-67 expression in adrenocortical carcinoma, Clinical Radiology, https://doi.org/10.1016/j.crad.2020.01.012
ROC Curve
1.0
Source of the Curve
Elongation
Flatness
0.8
Reference Line
Sensitivity
0.6
0.4-
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
1 - Specificity
linear regression model achieved moderate predictive ac- curacy for Ki-67. The independent predictors selected from the linear model included morphological features of the tumour, shape flatness, and elongation; morphological features provide a quantitative description of the physical appearance of the tumour, degree of solidity, surface ir- regularity, and eccentricity. Eccentricity described the de- gree of elongation of the tumour and its deviation from the regular round uniform circular appearance.32
The other significant predictors in the model were higher-order features that describe the intervoxel distri- bution within the image as GLCM-based feature (cluster shade), GRLRM (long run emphasis), GLSZM (grey-level non-uniformity), and NGTDM (contrast).33 Li et al. proposed a model involving nine radiomic features that had 83.3% accuracy for the prediction of Ki-67 expression in patients with low-grade glioma.30 Liang et al. established a T2- weighted image-derived radiomic score using 30 selected features to predict preoperative Ki-67 index in breast can- cer. The radiomic model exhibited good performance in differentiating high and low Ki-67 status, with an AUC of 0.7 and positive predictive value of 0.878.23
Further analysis of the independent predictors in the present model revealed that the AUC for the identification of high Ki-67 status was significantly higher for shape flatness and elongation than for other predictors in the present model. These two features showed significantly positive correlation with Ki-67 index by the Spearman rank method, which suggests that shape elongation and flatness could be superior to other features in identification of high
Ki-67 expression status in ACCs. A study on the correlation between dynamic contrast-enhanced MRI radiomic fea- tures and Ki-67 expression in invasive breast cancer found that standard deviation, skewness, kurtosis, contrast, ho- mogeneity, and inverse differential moment were signifi- cantly different between patients with a high Ki-67- expressing tumour and those with a low Ki-67-expressing tumour.31 A similar study to define the correlation be- tween CT-derived radiomic features and Ki-67 expression in lung cancer reported that inverse variance shape elongation and minor axis were promising radiomic features for the prediction of Ki-67 expression in these cancers.29
The present study has many limitations. First, the cohort was small (n=53 patients), which impeded the validation of the linear regression model. Second, it was a retrospective study. Third, the radiomic features could not be correlated with the clinical outcome because of differences in the follow-up periods among the patients and, in many cases, insufficient follow-up information. There were also multi- ple confounders that made assessment for recurrence difficult, including lack of information about resection margins in some tumours. Therefore, the correlation be- tween radiomic features and a well-established surrogate marker, Ki-67, was investigated instead. To avoid the vari- ability and the heterogeneity of the Ki-67 proliferation in- dex within different intratumour regions,9 the Ki-67 index was calculated from the entire tumour specimen instead of core biopsy samples.
In conclusion, this preliminary study indicates that a radiomic signature derived from contrast-enhanced CT
imaging might act as a non-invasive predictor of Ki-67 expression status in patients with ACC. Shape elongation and shape flatness were superior to other radiomic features in the differentiation between low and high Ki-67 expres- sion in the tumours. Future studies in larger cohorts are recommended to confirm the results in the present study.
Conflict of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper
References
1. Paragliola RM, Torino F, Papi G, et al. Role of mitotane in adrenocortical carcinoma - review and state of the art. Eur Endocrinol 2018;14:62-6.
2. Livhits M, Li N, Yeh MW, et al. Surgery is associated with improved survival for adrenocortical cancer, even in metastatic disease. Surgery 2014;156:1531-40.
3. Dickson PV, Kim L, Yen TWF, et al. Adjuvant and neoadjuvant therapy, treatment for advanced disease, and genetic considerations for adre- nocortical carcinoma: an update from the SSO Endocrine and Head and Neck Disease Site Working Group. Ann Surg Oncol 2018;25:3453-9.
4. Kerkhofs TM, Ettaieb MH, Hermsen IG, et al. Developing treatment for adrenocortical carcinoma. Endocr Relat Endocr Relat Canc 2015;22:R325-38.
5. Dudderidge TJ, Stoeber K, Loddo M, et al. Mcm2, Geminin, and KI-67 define proliferative state and are prognostic markers in renal cell car- cinoma. Clin Cancer Res 2005;11:2510-7.
6. Warth A, Cortis J, Soltermann A, et al. Tumour cell proliferation (Ki-67) in non-small cell lung cancer: a critical reappraisal of its prognostic role. Br J Cancer 2014;111:1222-9.
7. Beuschlein F, Weigel J, Saeger W, et al. Major prognostic role of Ki-67 in localized adrenocortical carcinoma after complete resection. J Clin Endocrinol Metab 2015;100:841-9.
8. Phan AT. Adrenal cortical carcinoma-review of current knowledge and treatment practices. Hematol Oncol Clin North Am 2007;21:489-507. viii-ix.
9. Fassnacht M, Dekkers O, Else T, et al. European Society of Endocrinology clinical practice guidelines on the management of adrenocortical car- cinoma in adults, in collaboration with the European Network for the Study of Adrenal Tumours. Eur J Endocrinol 2018, https://doi.org/ 10.1530/EJE-18-0608.
10. Davnall F, Yip CS, Ljungqvist G, et al. Assessment of tumour heteroge- neity: an emerging imaging tool for clinical practice? Insights Imaging 2012;3:573-89.
11. Parekh V, Jacobs MA. Radiomics: a new application from established techniques. Expert Rev Precis Med Drug Dev 2016;1:207-26.
12. Gillies RJ, Kinahan PE, Hricak H. Radiomics: images are more than pic- tures, they are data. Radiology 2016;278:563-77.
13. Alic L, Niessen WJ, Veenland JF. Quantification of heterogeneity as a biomarker in tumour imaging: a systematic review. PLoS One 2014;9:e110300, https://doi.org/10.1371/journal.pone.0110300.
14. Sala E, Mema E, Himoto Y, et al. Unravelling tumour heterogeneity using next-generation imaging: radiomics, radiogenomics, and habitat imag- ing. Clin Radiol 2017;72:3-10.
15. Boros M, Moncea D, Moldovan C, et al. Intratumoural heterogeneity for Ki-67 index in invasive breast carcinoma: a study on 131 consecutive cases. Appl Immunohistochem Mol Morphol 2017;25:338-40.
16. Li X, Morgan PS, Ashburner J, et al. The first step for neuroimaging data analysis: DICOM to NIfTI conversion. J Neurosci Methods 2016;264:47-56.
17. ThermoFischerScientific. Amira for life & biomedical sciences. 2018. Avail- able at: https://www.thermofisher.com/us/en/home/industrial/electron- microscopy/electron-microscopy-instruments-workflow-solutions/3d- visualization-analysis-software/amira-life-sciences-biomedical.html. [Accessed 15 April 2019].
18. van Griethuysen JM, Fedorov A, Parmar C, et al. Computational radio- mics system to decode the radiographic phenotype. Cancer Res 2017;77:e104-7. https://doi.org/10.1158/0008-5472.
19. IBM Corp. IBM SPSS statistics for windows, version 24.0. Available at: https://www.ibm.com/products/spss-statistics. [Accessed 23 July 2019].
20. Amin MB, Edge SB, editors. AJCC cancer staging manual. 8th edn. Cham, Switzerland: Springer; 2017.
21. Inwald EC, Klinkhammer-Schalke M, Hofstadter F, et al. Ki-67 is a prognostic parameter in breast cancer patients: results of a large population-based cohort of a cancer registry. Breast Canc Res Treat 2013;139:539-52.
22. Xie Y, Chen L, Ma X, et al. Prognostic and clinicopathological role of high Ki-67 expression in patients with renal cell carcinoma: a systematic review and meta-analysis. Sci Rep 2017;7:44281.
23. Liang C, Cheng Z, Huang Y, et al. An MRI-based radiomics classifier for preoperative prediction of Ki-67 status in breast cancer. Acad Radiol 2018;25:1111-7.
24. Libe R, Borget I, Ronchi CL, et al. Prognostic factors in stage III-IV adrenocortical carcinomas (ACC): an European Network for the Study of Adrenal Tumour (ENSAT) study. Ann Oncol 2015;26:2119-25.
25. Morimoto R, Satoh F, Murakami O, et al. Immunohistochemistry of a proliferation marker Ki-67/MIB1 in adrenocortical carcinomas: Ki-67/ MIB1 labeling index is a predictor for recurrence of adrenocortical carcinomas. Endocr J 2008;55:49-55.
26. Libé R. Adrenocortical carcinoma (ACC): diagnosis, prognosis, and treatment. Front Cell Dev Biol 2015;3:45. https://doi.org/10.3389/ fcell.2015.00045.
27. Crivelli P, Ledda RE, Parascandolo N, et al. A new challenge for radiolo- gists: radiomics in breast cancer. Biomed Res Int 2018;2018:6120703. https://doi.org/10.1155/2018/6120703.
28. Meyer HJ, Schob S, Hohn AK, et al. MRI texture analysis reflects histo- pathology parameters in thyroid cancer - a first preliminary study. Transl Oncol 2017;10:911-6.
29. Zhou B, Xu J, Tian Y, et al. Correlation between radiomic features based on contrast-enhanced computed tomography images and Ki-67 prolif- eration index in lung cancer: a preliminary study. Thorac Cancer 2018;9:1235-40.
30. Li Y, Qian Z, Xu K, et al. Radiomic features predict Ki-67 expression level and survival in lower grade gliomas. J Neurooncol 2017;135:317-24.
31. Juan MW, Yu J, Peng GX, et al. Correlation between DCE-MRI radiomics features and Ki-67 expression in invasive breast cancer. Oncol Lett 2018;16:5084-90.
32. Sutton EJ, Oh JH, Dashevsky BZ, et al. Breast cancer subtype intertumour heterogeneity: MRI-based features predict results of a genomic assay. J Magn Reson Imaging 2015;42:1398-406.
33. Parekh V, Jacobs MA. Radiomics: a new application from established techniques. Expert Rev Precis Med Drug Dev 2016;1:207-26.