Abstract
Objectives
To systematically review published studies on the use of radiomics of the pancreas.
Methods
The search was conducted in the MEDLINE database. Human studies that investigated the applications of radiomics in diseases of the pancreas were included. The radiomics quality score was calculated for each included study.
Results
A total of 72 studies encompassing 8863 participants were included. Of them, 66 investigated focal pancreatic lesions (pancreatic cancer, precancerous lesions, or benign lesions); 4, pancreatitis; and 2, diabetes mellitus. The principal applications of radiomics were differential diagnosis between various types of focal pancreatic lesions (n = 19), classification of pancreatic diseases (n = 23), and prediction of prognosis or treatment response (n = 30). Second-order texture features were most useful for the purpose of differential diagnosis of diseases of the pancreas (with 100% of studies investigating them found a statistically significant feature), whereas filtered image features were most useful for the purpose of classification of diseases of the pancreas and prediction of diseases of the pancreas (with 100% of studies investigating them found a statistically significant feature). The median radiomics quality score of the included studies was 28%, with the interquartile range of 22% to 36%. The radiomics quality score was significantly correlated with the number of extracted radiomics features (r = 0.52, p < 0.001) and the study sample size (r = 0.34, p = 0.003).
Conclusions
Radiomics of the pancreas holds promise as a quantitative imaging biomarker of both focal pancreatic lesions and diffuse changes of the pancreas. The usefulness of radiomics features may vary depending on the purpose of their application. Standardisation of image acquisition protocols and image pre-processing is warranted prior to considering the use of radiomics of the pancreas in routine clinical practice.
Key Points
• Methodologically sound studies on radiomics of the pancreas are characterised by a large sample size and a large number of extracted features.
• Optimisation of the radiomics pipeline will increase the clinical utility of mineable pancreas imaging data.
• Radiomics of the pancreas is a promising personalised medicine tool in diseases of the pancreas.
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Avoid common mistakes on your manuscript.
Introduction
Diseases of the pancreas are complex (with a wide array of genetic, environmental, and behavioural factors affecting them) and often lie on a continuum. Acute pancreatitis (AP) is the most common disease of the exocrine pancreas with the global incidence of 33.7 per 100,000 individuals per year [1]. One-fifth of individuals after first episode of AP develop recurrent acute pancreatitis (RAP), and 36% of those with RAP progress to chronic pancreatitis (CP) [2]. Pancreatic cancer is the most lethal disease of the pancreas with the global incidence and mortality of 8.1 and 6.9 per 100,000 general population per year, respectively [1]. Its common risk factors include familial pancreatic cancer kindred and deleterious germline mutations in pancreatic cancer susceptibility genes [3]. Also, several focal pancreatic lesions (pancreatic intraepithelial neoplasms-grade 3 (PanIN-3), intraductal papillary mucinous neoplasms, and mucinous cystic neoplasms) are considered precancerous [4]. Both pancreatitis and pancreatic cancer often lead to new-onset diabetes, termed ‘diabetes of the exocrine pancreas’—the second most common type of new-onset diabetes in adults [5].
Imaging modalities (such as computed tomography (CT), magnetic resonance imaging (MRI), endoscopic ultrasonography, and positron emission tomography) are frequently used in management of diseases of the pancreas [6, 7]. Traditionally, their use predominantly includes subjective assessment of a handful of generic qualitative features that describe the underlying pathology of the pancreas. However, images of the pancreas contain an innumerable amount of objective data specific to each patient that could be harnessed to provide personalised management of patients [8, 9]. The field of quantitative image analysis has evolved in recent years and automatedly extracted features can now be analysed. The process of high-throughput extraction of image features from radiological images has been termed ‘radiomics’ [10]. Organ-specific radiomics promises to be a cornerstone of personalised medicine in the future. The use of radiomics in lung, liver, prostate, breast, kidney, rectum, and central nervous system diseases has been reviewed [11,12,13,14,15,16,17]. However, to date, there has been no systematic review on the use of radiomics in diseases of the pancreas.
The aim was to systematically benchmark published studies on radiomics of the pancreas and to determine their quality as well as the factors that are associated with it.
Methods
Search strategy
The search strategy was conducted in consultation with an experienced subject librarian to identify all relevant studies that reported on the use of radiomics of the pancreas in humans. A systematic literature search was conducted to identify all studies published from January 1, 2000 to April 15, 2020, using the MEDLINE database. No language restrictions were applied. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed. The initial screening was done through the review of titles and abstracts. Full-text articles of potentially relevant studies were retrieved and assessed for eligibility. Relevant articles were also identified through reference lists of the retrieved full-text articles.
Eligibility criteria
Eligible studies had to investigate the applications of image analysis in pancreatic benign or precancerous lesions, pancreatic cancer, pancreatitis, or diabetes mellitus through extracting quantitative imaging features (i.e. radiomics). All imaging modalities were eligible. Studies were excluded if they were conducted not in humans; qualitative imaging features alone were reported; machine learning (e.g. convolutional neural network) was used to recognise image patterns without extracting quantitative features; they focused on technical (e.g. image pre-processing or image acquisition parameters) or patient-related parameters and their effect on the stability and reproducibility of extracted features; and they focused on a complication of pancreatic surgery (e.g. pancreatic fistula). Publications other than original articles (e.g. reviews, book chapters, editorials) were not considered.
Data extraction
The following data were extracted from the included studies, if available: authors, year of publication, country, cohort size, goal(s) of the study, type of imaging modality, parameters of imaging (e.g. slice thickness, imaging phase, MRI sequence), method of segmentation, type of feature extraction software, type of extracted features (i.e. quantitative radiomics only or semantic), type and number of extracted quantitative features, number of statistically significant quantitative features, and method of feature reduction and classification.
The included studies were grouped into three main categories based on the main goal of each study. The first category was differential diagnosis of diseases of the pancreas (e.g. differentiation between healthy pancreas and chronic pancreatitis or pancreatic cancer). The second category was classification of diseases of the pancreas, where more than two subtypes of the same disease of the pancreas were studied (e.g. classification of subtypes of pancreatitis or classification of histologic grades of pancreatic cancer). The third category was prediction of diseases of the pancreas (e.g. prediction of survival in patients with unresectable pancreatic cancer or prediction of patient response to a certain treatment). Within each category, all radiomics features were categorised into three main groups (at least one significant feature reported, no significant feature reported, and non-investigated feature) with the view to determining a clinically useful pattern.
Radiomics quality score
The radiomics quality score (RQS) was calculated for each individual study. In brief, the RQS assessed the quality of radiomics study in terms of robustness and reproducibility through assigning points based on 16 criteria [18]. The number of points depended on the importance of the respective criterion, with 36 points (100%) being the maximum number.
Statistical analysis
The associations of the RQS with number of extracted features and cohort size were investigated using Pearson correlation coefficient. The association between the RQS and the type of imaging modality was investigated using linear regression analysis (with CT set as the reference). Statistical analysis was performed using SPSS software (version 24). A p value of < 0.05 was considered statistically significant.
Results
Characteristics of the included studies
The total number of retrieved publications was 120 (Fig. 1). Seventy-two studies met the eligibility criteria and were included in the systematic literature review [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90]. These studies encompassed a total of 8863 individuals. The sample size varied between 17 and 690 individuals, with a median of 100 individuals. Fifty-four studies (75.0%) employed CT [23, 24, 26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43, 45,46,47,48,49, 51, 52, 55,56,57, 59, 61,62,63,64,65, 68,69,70,71, 74,75,76,77, 80,81,82,83, 85,86,87,88,89,90]; nine studies (12.5%), MRI [44, 50, 53, 54, 58, 60, 66, 67, 78]; three studies (4.2%), endoscopic ultrasound [19,20,21]; and six studies (8.3%), positron emission tomography [22, 25, 72, 73, 79, 84]. Forty-four studies (61.1%) were conducted in Asia [19,20,21, 33, 35,36,37,38, 41,42,43,44, 48,49,50,51,52, 56,57,58,59,60,61, 63, 64, 66,67,68,69,70,71,72,73,74, 76, 77, 79, 80, 83, 84, 87,88,89,90], 20 studies (27.8%) in North America [22,23,24,25,26,27,28,29,30,31,32, 39, 40, 45, 55, 62, 65, 75, 82, 85], and eight studies (11.1%) in Europe [34, 46, 47, 53, 54, 78, 81, 86]. Other details are presented in Table 1.
Sixty-six (91.7%) studies applied manual segmentation of the pancreas, whereas six studies used semi-automated segmentation [24, 25, 29, 46, 56, 84]. Twenty-seven studies (37.5%) extracted only quantitative radiomics features, whereas 45 studies combined both radiomics and semantic features. The semantic features in the 45 studies were clinical features (e.g. age, gender, body mass index (n = 45), histopathological features (e.g. tumour grades, mitotic index) (n = 15), blood biomarkers (e.g. cancer antigen 19-9, carcinoembryonic antigen) (n = 14), and genetic signatures (e.g. HMGA2 and c-Myc genes, miRNA genomic classifier) (n = 2). Out of the 45 studies, 20 studies reported that the performance of combined model (i.e. significant radiomics features plus semantic features) is higher than the performance of radiomics model alone. The superiority of combined model was confirmed statistically in eight studies [22, 24, 31, 46, 48, 51, 59, 88].
Various approaches to dimensionality reduction were applied in the included studies in order to select the most useful radiomics feature and reduce the effect of overfitting. These approaches included univariate filter technique (n = 17), multivariate filter technique (n = 20), least absolute shrinkage and selection operator regression (n = 18), as well as principle component analysis (n = 2). The useful features were used as an input for training and validating classification model. Out of the 72 included studies, 36 studies applied supervised machine learning techniques (including random forest in nine studies); 14 studies, support vector machine; and 22 studies, logistic regression. The median RQS of the included studies was 28% (interquartile range 22–36%). The three most frequently observed RQS characteristics were discrimination statistics and applying resampling techniques, employing well-documented imaging protocol, and clinical usefulness of the model (Fig. 2). The three least frequently observed RQS characteristics were prospective study design, imaging at different time points, and comparing radiomics model with current gold standard method (Fig. 2). Fourteen studies (19.5%) analysed feature robustness through detecting inter-scanner differences and vendor-dependent features [33, 38, 42, 48, 55, 58, 61,62,63, 69, 71, 73, 82, 86].
Applications of radiomics of the pancreas
The median number of extracted radiomics features in the included studies was 166 (interquartile range 14–416). Eighteen studies (25%) used in-house developed software [22, 24, 28,29,30, 33, 36, 38, 39, 45, 56, 61, 63, 69, 72, 80, 85, 90], whereas the remaining studies used open source or commercial software. The types and number of extracted features in individual studies are presented in Table 2. The significant radiomics features in individual studies are presented in Tables 3, 4, and 5 (stratified by the primary goal of using radiomics). The main focus of radiomics in 56 studies (77.8%) was pancreatic cancer; four studies (5.6%), pancreatic precancerous lesions [23, 24, 32, 39]; six studies (8.4%), pancreatic benign lesions [29, 68,69,70, 87, 88]; four studies (5.6%), pancreatitis [42, 60, 78, 85]; and two studies (2.8%), diabetes mellitus [61, 80]. Nineteen studies (26.4%) primarily applied radiomics for differentiation between various diseases of the pancreas, 23 studies (32.0%) for classification of subtypes/histologic grades, and 30 studies (41.7%) for prediction of prognosis/treatment response (Fig. 3). Out of the 72 included studies, 28 studies extracted radiomics features and patterns with the use of different filters (including wavelet, square, square root, exponential, logarithm, gradient, Laws, local binary pattern, Laplacian of Gaussian, and fractal dimension filters).
Factors that affect radiomics quality score
Supplementary Table 1 details the RQS of individual studies. Overall, the RQS was significantly correlated with the number of extracted features (r = 0.529, p < 0.001) as well as with the cohort size (r = 0.343, p = 0.003). In its turn, the number of extracted features and the cohort size were significantly correlated with the number of statistically significant features (r = 0.437, p < 0.001; and r = 0.437, p < 0.001, correspondingly). Using CT as the reference, radiomics features extracted from MRI images resulted in an increase in the RQS by 1 point (p = 0.732); extracted from positron emission tomography images, in an increase in the RQS by 2 points (p = 0.570); extracted from endoscopic ultrasonography images, in a decrease in the RQS by 10 points (p = 0.102). Using CT as the reference, radiomics features extracted from MRI images resulted in 5 more statistically significant features (p = 0.134); extracted from endoscopic ultrasonography images, in 3 more statistically significant features (p = 0.495); extracted from positron emission tomography images, in 1 less statistically significant feature (p = 0.680).
Discussion
This is the first systematic review to investigate the use of radiomics of the pancreas and the factors that affect quantitative imaging features of the pancreas. A total of 72 studies that enrolled more than eight thousand participants were included, with the median sample size of 100 participants. The median number of investigated radiomics features in the included studies was 166, ranging from 4 to 2041 features. These features could be grouped into five main categories: shape features, first-order texture features, second-order texture features, filtered image features, and customised features [91]. Filtered image features appeared to be the most frequently observed significant radiomics features in the studies that employed them for classification (Table 4) and prediction (Table 5) of diseases of the pancreas. However, only 39% of studies (28 out of 72) used these features and more research is needed to confirm their usefulness in diseases of the pancreas. Future research also needs to determine the optimal filters as the included studies used a total of 10 different filters (including wavelet, square, square root, exponential, logarithm, gradient, Laws, local binary pattern, Laplacian of Gaussian, and fractal dimension filters). Second-order texture features were used in 94% of studies (68 out of 72) and they appeared to be the most frequently observed significant radiomics features in the studies focused on differential diagnosis of diseases of the pancreas (Table 3). The superiority of this group of features is likely explained by the fact that they capture the spatial arrangement and distribution of intensities within the pancreas using different types of matrices (e.g. grey level co-occurrence matrix, grey level run length matrix). Making use of large number of matrices may be required as the pancreas is a complex glandular organ of a relatively small size, located deeply in the retroperitoneal space, and composed of different types of cells (i.e. acinar, ductal, endocrine)—each with different functions [92]. The exocrine part constitutes around 95% of the pancreas with two main types of cells being acinar cells and ductal cells. The endocrine part (i.e. islets of Langerhans) constitutes less than 5% of the pancreas, with five major types of cells being alpha cells, beta cells, delta cells, epsilon cells, and pancreatic polypeptide cells. Besides, the size of the pancreas may change during consumption of food [93,94,95]. These physical and physiological characteristics of the pancreas make radiomics investigation of the pancreas quite challenging and justify the extraction of large number of radiomics features that could describe the pancreas comprehensively. Further, extracting large number of features likely captures variabilities in genetic, environmental, and behavioural factors that cause diseases of the pancreas, hence enabling characterisation of each patient individually and ultimately resulting in personalised management [1]. In the future, the performance of radiomics models may, in principle, be enhanced by considering certain semantic features (e.g. demographics, blood biomarkers, genomics). However, it is a long way to go as, out of the 45 studies that reported on combined models, only 8 studies (18%) demonstrated a statistically significant superiority of combining radiomics and semantic features.
The other notable finding of the present systematic review was that the RQS had a significant positive correlation with the number of extracted features and the cohort size. However, the RQS in all the included studies altogether was rather low, with a median of 28%. The top three most consistently reported criteria were reporting discrimination statistics, applying well-documented imaging protocol, and studying the clinical utility of the extracted biomarker. By contrast, the three least frequently reported criteria were prospective study design, applying the delta radiomics (i.e. extracting features at different imaging time points), and comparing the results with gold standard. Worryingly, none of the included studies were prospective. Therefore, future radiomics studies of the pancreas should be conducted in a prospective fashion. The second least frequently observed criterion was the use of delta radiomics, where multiple images were obtained at different time points in order to test the reproducibility and stability of extracted radiomics features over a specific period of time. Out of the 72 included studies, only three studies met this criterion. Therefore, future studies on radiomics of the pancreas should extract and test radiomics features at multiple time points. The third common omission was the lack of comparison of radiomics findings with gold standard. For example, results of many studies on the use of radiomics of the pancreas to determine prognosis of patients with pancreatic cancer were not compared with the well-established gold standards (tumour node metastasis (TNM) staging system and MD-Anderson pre-treatment classification) [96]. It is also worth noting that, while histology is the gold standard for diagnosing focal pancreatic lesions, only 15 included studies used it (although it may not be ethical to use it in patients with benign lesions). Further, all the 6 radiomics studies on pancreatic benign lesions (such as serous cystadenoma) used CT only, which is considered suboptimal as accurate diagnosing of these benign lesions is quite challenging without the use of MRI (especially if lesions are of small size). Careful selection of gold standard in future studies on radiomics of the pancreas is encouraged.
There are several limitations that need to be acknowledged when interpreting the findings of the present review. First, there was a heterogeneity between the included studies in terms of image acquisition protocols. For example, different phases of CT and MRI sequences were employed in the primary studies. This brings to the fore the need to standardise image acquisition protocols in future radiomics studies of the pancreas. Second, the included studies used a range of software packages that not infrequently offer different algorithms for defining the same radiomics features. This highlights the need to standardise the definitions of imaging features. The Image Biomarker Standardisation Initiative aspires to standardise the extraction of imaging biomarkers from acquired imaging for the purpose of high-throughput quantitative image analysis [97]. This initiative needs to be taken into account in radiomics of the pancreas research. Third, the included studies disproportionately focused on focal pancreatic lesions. Only six studies investigated benign diseases that are characterised by diffuse changes of the pancreas (i.e. pancreatitis and diabetes mellitus) and high-quality radiomics studies in these diseases are now warranted [98]. Fourth, we designed the present systematic review to include only studies that applied handcrafted radiomics as the main method for extracting quantitative imaging features from radiological images. However, it is also possible to use machine learning to extract some features. For example, one study extracted 256 deep learning features from the first three layers of convolutional neural network model, in addition to the radiomics features [57]. Last, building predictive model is one of the promising applications of radiomics in diseases of the pancreas. Thirty studies applied radiomics for building predictive models; however, none of them appeared to follow the TRIPOD (Transparent Reporting of multivariable prediction model for Individual Prognosis or Diagnosis) guidelines [99].
In conclusion, the present systematic review demonstrated that radiomics of the pancreas emerges as a promising tool that could be used for personalised management of patients with diseases of the pancreas. To maximise the benefits of radiomics of the pancreas, future studies are best to have a large sample size (more than 100 participants), use standardised software packages that offer a large number of radiomics features (especially second-order texture features and filtered image features), investigate radiomics in prospective fashion, compare radiomics results with an appropriate gold standard, and apply delta radiomics.
Abbreviations
- RQS:
-
Radiomics quality score
References
Xiao AY, Tan MLY, Wu LM et al (2016) Global incidence and mortality of pancreatic diseases: a systematic review, meta-analysis, and meta-regression of population-based cohort studies. Lancet Gastroenterol Hepatol 1:45–55
Sankaran SJ, Xiao AY, Wu LM, Windsor JA, Forsmark CE, Petrov MS (2015) Frequency of progression from acute to chronic pancreatitis and risk factors: a meta-analysis. Gastroenterology 149:1490–1500
Becker AE, Hernandez YG, Frucht H, Lucas AL (2014) Pancreatic ductal adenocarcinoma: risk factors, screening, and early detection. World J Gastroenterol 20:11182–11198
Hruban RH, Takaori K, Klimstra DS et al (2004) An illustrated consensus on the classification of pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms. Am J Surg Pathol 28:977–987
Petrov MS (2017) Diabetes of the exocrine pancreas: American Diabetes Association-compliant lexicon. Pancreatology 17:523–526
Kumar H, DeSouza SV, Petrov MS (2019) Automated pancreas segmentation from computed tomography and magnetic resonance images: a systematic review. Comput Methods Programs Biomed 178:319–328
DeSouza SV, Priya S, Cho J, Singh RG, Petrov MS (2019) Pancreas shrinkage following recurrent acute pancreatitis: an MRI study. Eur Radiol 29:3746–3756
Gillies RJ, Kinahan PE, Hricak H (2016) Radiomics: images are more than pictures, they are data. Radiology 278:563–577
Petrov MS (2018) Harnessing analytic morphomics for early detection of pancreatic cancer. Pancreas 47:1051–1054
Kumar V, Gu Y, Basu S et al (2012) Radiomics: the process and the challenges. Magn Reson Imaging 30:1234–1248
Thawani R, McLane M, Beig N et al (2018) Radiomics and radiogenomics in lung cancer: a review for the clinician. Lung Cancer 115:34–41
Fan Y, Feng M, Wang R (2019) Application of radiomics in central nervous system diseases: a systematic literature review. Clin Neurol Neurosurg 187:105565
Wakabayashi T, Ouhmich F, Gonzalez-Cabrera C et al (2019) Radiomics in hepatocellular carcinoma: a quantitative review. Hepatol Int 13:546–559
Sun Y, Reynolds HM, Parameswaran B et al (2019) Multiparametric MRI and radiomics in prostate cancer: a review. Australas Phys Eng Sci Med 42:3–25
Valdora F, Houssami N, Rossi F, Calabrese M, Tagliafico AS (2018) Rapid review: radiomics and breast cancer. Breast Cancer Res Treat 169:217–229
Chen J, Remulla D, Nguyen JH et al (2019) Current status of artificial intelligence applications in urology and their potential to influence clinical practice. BJU Int 124:567–577
Horvat N, Bates DD, Petkovska I (2019) Novel imaging techniques of rectal cancer: what do radiomics and radiogenomics have to offer? A literature review. Abdom Radiol (NY) 44:3764–3774
Lambin P, Leijenaar RT, Deist TM et al (2017) Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 14:749–762
Zhang M-M, Yang H, Jin Z-D, Yu J-G, Cai Z-Y, Li Z-S (2010) Differential diagnosis of pancreatic cancer from normal tissue with digital imaging processing and pattern recognition based on a support vector machine of EUS images. Gastrointest Endosc 72:978–985
Xu W, Liu Y, Lu Z et al (2013) A new endoscopic ultrasonography image processing method to evaluate the prognosis for pancreatic cancer treated with interstitial brachytherapy. World J Gastroenterol 19:6479–6484
Zhu M, Xu C, Yu J et al (2013) Differentiation of pancreatic cancer and chronic pancreatitis using computer-aided diagnosis of endoscopic ultrasound (EUS) images: a diagnostic test. PLoS One 8:e63820
Cui Y, Song J, Pollom E et al (2016) Quantitative analysis of (18)F-fluorodeoxyglucose positron emission tomography identifies novel prognostic imaging biomarkers in locally advanced pancreatic cancer patients treated with stereotactic body radiation therapy. Int J Radiat Oncol Biol Phys 96:102–109
Hanania AN, Bantis LE, Feng Z et al (2016) Quantitative imaging to evaluate malignant potential of IPMNs. Oncotarget 7:85776–85784
Permuth JB, Choi J, Balarunathan Y et al (2016) Combining radiomic features with a miRNA classifier may improve prediction of malignant pathology for pancreatic intraductal papillary mucinous neoplasms. Oncotarget 7:85785–85797
Yue Y, Osipov A, Fraass B et al (2016) Identifying prognostic intratumor heterogeneity using pre- and post-radiotherapy 18F-FDG PET images for pancreatic cancer patients. J Gastrointest Oncol 8:127–138
Canellas R, Burk KS, Parakh A, Sahani DV (2017) Prediction of pancreatic neuroendocrine tumor grade based on CT features and texture analysis. AJR Am J Roentgenol 210:341–346
Cassinotto C, Chong J, Zogopoulos G et al (2017) Resectable pancreatic adenocarcinoma: role of CT quantitative imaging biomarkers for predicting pathology and patient outcomes. Eur J Radiol 90:152–158
Chen X, Oshima K, Schott D et al (2017) Assessment of treatment response during chemoradiation therapy for pancreatic cancer based on quantitative radiomic analysis of daily CTs: an exploratory study. PLoS One 12:e0178961
Dmitriev K, Kaufman AE, Javed AA et al (2017) Classification of pancreatic cysts in computed tomography images using a random forest and convolutional neural network ensemble. Med Image Comput Comput Assist Interv 10435:150–158
Eilaghi A, Baig S, Zhang Y et al (2017) CT texture features are associated with overall survival in pancreatic ductal adenocarcinoma – a quantitative analysis. BMC Med Imaging 17:38
Attiyeh MA, Chakraborty J, Doussot A et al (2018) Survival prediction in pancreatic ductal adenocarcinoma by quantitative computed tomography image analysis. Ann Surg Oncol 25:1034–1042
Chakraborty J, Midya A, Gazit L et al (2018) CT radiomics to predict high-risk intraductal papillary mucinous neoplasms of the pancreas. Med Phys 45:5019–5029
Choi TW, Kim JH, Yu MH, Park SJ, Han JK (2018) Pancreatic neuroendocrine tumor: prediction of the tumor grade using CT findings and computerized texture analysis. Acta Radiol 59:383–392
Ciaravino V, Cardobi N, de Robertis R et al (2018) CT texture analysis of ductal adenocarcinoma downstaged after chemotherapy. Anticancer Res 38:4889–4895
Guo C, Zhuge X, Wang Q et al (2018) The differentiation of pancreatic neuroendocrine carcinoma from pancreatic ductal adenocarcinoma: the values of CT imaging features and texture analysis. Cancer Imaging 18:37
Li J, Lu J, Liang P et al (2018) Differentiation of atypical pancreatic neuroendocrine tumors from pancreatic ductal adenocarcinomas: using whole-tumor CT texture analysis as quantitative biomarkers. Cancer Med 7:4924–4931
Lin X, Xu L, Wu A, Guo C, Chen X, Wang Z (2018) Differentiation of intrapancreatic accessory spleen from small hypervascular neuroendocrine tumor of the pancreas: textural analysis on contrast-enhanced computed tomography. Acta Radiol 60:553–560
Yun G, Kim YH, Lee YJ, Kim B, Hwang J-H, Choi DJ (2018) Tumor heterogeneity of pancreas head cancer assessed by CT texture analysis: association with survival outcomes after curative resection. Sci Rep 8:7226
Attiyeh MA, Chakraborty J, Gazit L et al (2019) Preoperative risk prediction for intraductal papillary mucinous neoplasms by quantitative CT image analysis. HPB (Oxford) 21:212–218
Attiyeh MA, Chakraborty J, McIntyre CA et al (2019) CT radiomics associations with genotype and stromal content in pancreatic ductal adenocarcinoma. Abdom Radiol (NY) 44:3148–3157
Bian Y, Guo S, Jiang H et al (2019) Relationship between radiomics and risk of lymph node metastasis in pancreatic ductal adenocarcinoma. Pancreas 48:1195–1203
Chen Y, T-w C, Wu C-q et al (2019) Radiomics model of contrast-enhanced computed tomography for predicting the recurrence of acute pancreatitis. Eur Radiol 29:4408–4417
Cheng S-H, Cheng Y-J, Jin Z-Y, Xue H-D (2019) Unresectable pancreatic ductal adenocarcinoma: role of CT quantitative imaging biomarkers for predicting outcomes of patients treated with chemotherapy. Eur J Radiol 113:188–197
Choi MH, Lee YJ, Yoon SB, Choi J-I, Jung SE, Rha SE (2019) MRI of pancreatic ductal adenocarcinoma: texture analysis of T2-weighted images for predicting long-term outcome. Abdom Radiol (NY) 44:122–130
Chu LC, Park S, Kawamoto S et al (2019) Utility of CT radiomics features in differentiation of pancreatic ductal adenocarcinoma from normal pancreatic tissue. AJR Am J Roentgenol 213:349–357
Cozzi L, Comito T, Fogliata A et al (2019) Computed tomography based radiomic signature as predictive of survival and local control after stereotactic body radiation therapy in pancreatic carcinoma. PLoS One 14:e0210758
D’Onofrio M, Ciaravino V, Cardobi N et al (2019) CT enhancement and 3D texture analysis of pancreatic neuroendocrine neoplasms. Sci Rep 9:2176
Gu D, Hu Y, Ding H et al (2019) CT radiomics may predict the grade of pancreatic neuroendocrine tumors: a multicenter study. Eur Radiol 29:6880–6890
Guo C, Zhuge X, Wang Z et al (2019) Textural analysis on contrast-enhanced CT in pancreatic neuroendocrine neoplasms: association with WHO grade. Abdom Radiol (NY) 44:576–585
Guo C-g, Ren S, Chen X et al (2019) Pancreatic neuroendocrine tumor: prediction of the tumor grade using magnetic resonance imaging findings and texture analysis with 3-T magnetic resonance. Cancer Manag Res 11:1933–1944
He M, Liu Z, Lin Y et al (2019) Differentiation of atypical non-functional pancreatic neuroendocrine tumor and pancreatic ductal adenocarcinoma using CT based radiomics. Eur J Radiol 117:102–111
Huang Z, Li M, He D et al (2019) Two-dimensional texture analysis based on CT images to differentiate pancreatic lymphoma and pancreatic adenocarcinoma: a preliminary study. Acad Radiol 26:e189–e195
Kaissis G, Ziegelmayer S, Lohöfer F et al (2019) A machine learning model for the prediction of survival and tumor subtype in pancreatic ductal adenocarcinoma from preoperative diffusion-weighted imaging. Eur Radiol Exp 3:41
Kaissis G, Ziegelmayer S, Lohöfer F et al (2019) A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy. PLoS One 14:e0218642
Khalvati F, Zhang Y, Baig S et al (2019) Prognostic value of CT radiomic features in resectable pancreatic ductal adenocarcinoma. Sci Rep 9:5449
Kim HS, Kim YJ, Kim KG, Park JS (2019) Preoperative CT texture features predict prognosis after curative resection in pancreatic cancer. Sci Rep 9:17389
Li K, Xiao J, Yang J et al (2019) Association of radiomic imaging features and gene expression profile as prognostic factors in pancreatic ductal adenocarcinoma. Am J Transl Res 11:4491–4499
Li X, Zhu H, Qian X, Chen N, Lin X (2020) MRI texture analysis for differentiating nonfunctional pancreatic neuroendocrine neoplasms from solid pseudopapillary neoplasms of the pancreas. Acad Radiol 27:815–823
Liang W, Yang P, Huang R et al (2019) A combined nomogram model to preoperatively predict histologic grade in pancreatic neuroendocrine tumors. Clin Cancer Res 25:584–594
Lin Q, Y-f JI, Chen Y et al (2019) Radiomics model of contrast-enhanced MRI for early prediction of acute pancreatitis severity. J Magn Reson Imaging 51:397–406
Lu C-Q, Wang Y-C, Meng X-P et al (2019) Diabetes risk assessment with imaging: a radiomics study of abdominal CT. Eur Radiol 29:2233–2242
Nasief H, Zheng C, Schott D et al (2019) A machine learning based delta-radiomics process for early prediction of treatment response of pancreatic cancer. NPJ Precis Oncol 3:25
Qiu W, Duan N, Chen X et al (2019) Pancreatic ductal adenocarcinoma: machine learning–based quantitative computed tomography texture analysis for prediction of histopathological grade. Cancer Manag Res 11:9253–9264
Ren S, Zhang J, Chen J et al (2019) Evaluation of texture analysis for the differential diagnosis of mass-forming pancreatitis from pancreatic ductal adenocarcinoma on contrast-enhanced CT images. Front Oncol 9:1171
Sandrasegaran K, Lin Y, Asare-Sawiri M, Taiyini T, Tann M (2019) CT texture analysis of pancreatic cancer. Eur Radiol 29:1067–1073
Tang TY, Li X, Zhang Q et al (2020) Development of a novel multiparametric MRI radiomic nomogram for preoperative evaluation of early recurrence in resectable pancreatic cancer. J Magn Reson Imaging 52:231–245
Wang YW, Zhang XH, Wang BT et al (2019) Value of texture analysis of intravoxel incoherent motion parameters in differential diagnosis of pancreatic neuroendocrine tumor and pancreatic adenocarcinoma. Chin Med Sci J 34:1–9
Wei R, Lin K, Guo Y, Li J, Wang Y (2019) Feasibility analysis of predicting expression of Ki67 in pancreatic cystic neoplasm based on radiomics. J Biomed Eng 36:1–6
Wei R, Lin K, Yan W et al (2019) Computer-aided diagnosis of pancreas serous cystic neoplasms: a radiomics method on preoperative MDCT images. Technol Cancer Res Treat 18:1533033818824339
Yang J, Guo X, Ou X, Zhang W, Ma X (2019) Discrimination of pancreatic serous cystadenomas from mucinous cystadenomas with CT textural features: based on machine learning. Front Oncol 9:494
Yu H, Huang Z, Li M et al (2019) Differential diagnosis of nonhypervascular pancreatic neuroendocrine neoplasms from pancreatic ductal adenocarcinomas, based on computed tomography radiological features and texture analysis. Acad Radiol 3:332–341
Zhang Y, Cheng C, Liu Z et al (2019) Differentiation of autoimmune pancreatitis and pancreatic ductal adenocarcinoma based on multi-modality texture features in 18F-FDG PET/CT. Sheng Wu Yi Xue Gong Cheng Xue Za Zhi 36:755–762
Zhang Y, Cheng C, Liu Z et al (2019) Radiomics analysis for the differentiation of autoimmune pancreatitis and pancreatic ductal adenocarcinoma in 18F-FDG PET/CT. Med Phys 46:4520–4530
Zhou HF, Han YQ, Lu J et al (2019) Radiomics facilitates candidate selection for irradiation stents among patients with unresectable pancreatic cancer. Front Oncol 9:973
Borhani AA, Dewan R, Furlan A et al (2020) Assessment of response to neoadjuvant therapy using CT texture analysis in patients with resectable and borderline resectable pancreatic ductal adenocarcinoma. AJR Am J Roentgenol 214:362–369
Chang N, Cui L, Luo Y, Chang Z, Yu B, Liu Z (2020) Development and multicenter validation of a CT-based radiomics signature for discriminating histological grades of pancreatic ductal adenocarcinoma. Quant Imaging Med Surg 10:692–702
Fang WH, Li XD, Zhu H et al (2020) Resectable pancreatic ductal adenocarcinoma: association between preoperative CT texture features and metastatic nodal involvement. Cancer Imaging 20:17
Frøkjær JB, Lisitskaya MV, Jørgensen AS et al (2020) Pancreatic magnetic resonance imaging texture analysis in chronic pancreatitis: a feasibility and validation study. Abdom Radiol (NY) 5:1497–1506
Gao J, Huang X, Meng H et al (2020) Performance of multiparametric functional imaging and texture analysis in predicting synchronous metastatic disease in pancreatic ductal adenocarcinoma patients by hybrid PET/MR: initial experience. Front Oncol 10:198
Jang S, Kim JH, Choi S-Y, Park SJ, Han JK (2020) Application of computerized 3D-CT texture analysis of pancreas for the assessment of patients with diabetes. PLoS One 15:e0227492
Kaissis GA, Ziegelmayer S, Lohöfer FK et al (2020) Image-based molecular phenotyping of pancreatic ductal adenocarcinoma. J Clin Med 9:724
Kulkarni A, Carrion-Martinez I, Jiang NN et al (2020) Hypovascular pancreas head adenocarcinoma: CT texture analysis for assessment of resection margin status and high-risk features. Eur Radiol 30:2853–2860
Li K, Yao Q, Xiao J et al (2020) Contrast-enhanced CT radiomics for predicting lymph node metastasis in pancreatic ductal adenocarcinoma: a pilot study. Cancer Imaging 20:12
Lim CH, Cho YS, Choi JY et al (2020) Imaging phenotype using 18 F-fluorodeoxyglucose positron emission tomography–based radiomics and genetic alterations of pancreatic ductal adenocarcinoma. Eur J Nucl Med Mol Imaging 47:2113–2122
Mashayekhi R, Parekh VS, Faghih M, Singh VK, Jacobs MA, Zaheer A (2020) Radiomic features of the pancreas on CT imaging accurately differentiate functional abdominal pain, recurrent acute pancreatitis, and chronic pancreatitis. Eur J Radiol 123:108778
Reinert CP, Baumgartner K, Hepp T, Bitzer M, Horger M (2020) Complementary role of computed tomography texture analysis for differentiation of pancreatic ductal adenocarcinoma from pancreatic neuroendocrine tumors in the portal-venous enhancement phase. Abdom Radiol (NY) 45:750–758
Shen X, Yang F, Yang P et al (2020) A contrast-enhanced computed tomography based radiomics approach for preoperative differentiation of pancreatic cystic neoplasm subtypes: a feasibility study. Front Oncol 10:248
Xie T, Wang X, Li M, Tong T, Yu X, Zhou Z (2020) Pancreatic ductal adenocarcinoma: a radiomics nomogram outperforms clinical model and TNM staging for survival estimation after curative resection. Eur Radiol 30:2513–2524
Xie T, Wang X, Li M, Tong T, Yu X, Zhou Z (2020) Pancreatic ductal adenocarcinoma: a radiomics nomogram outperforms clinical model and TNM staging for survival estimation after curative resection. Eur Radiol 30:2513–2524
Zhao Z, Bian Y, Jiang H et al (2020) CT-radiomic approach to predict G1/2 nonfunctional pancreatic neuroendocrine tumor. Acad Radiol. https://doi.org/10.1016/j.acra.2020.01.002
Larue RT, Defraene G, De Ruysscher D, Lambin P, Van Elmpt W (2017) Quantitative radiomics studies for tissue characterization: a review of technology and methodological procedures. Br J Radiol 90:20160665
Das SL, Kennedy JI, Murphy R, Phillips AR, Windsor JA, Petrov MS (2014) Relationship between the exocrine and endocrine pancreas after acute pancreatitis. World J Gastroenterol 45:17196–17205
Pendharkar SA, Asrani VM, Xiao AY et al (2016) Relationship between pancreatic hormones and glucose metabolism: a cross-sectional study in patients after acute pancreatitis. Am J Physiol Gastrointest Liver Physiol 311:G50–58
Desouza SV, Yoon HD, Singh RG, Petrov MS (2018) Quantitative determination of pancreas size using anatomical landmarks and its clinical relevance: a systematic literature review. Clin Anat 31:913–926
DeSouza SV, Singh RG, Yoon HD, Murphy R, Plank LD, Petrov MS (2018) Pancreas volume in health and disease: a systematic review and meta-analysis. Expert Rev Gastroenterol Hepatol 12:757–766
Edge SB, Compton CC (2010) The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 17:1471–1474
Zwanenburg A, Vallières M, Abdalah MA et al (2020) The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology 295:328–338
Pendharkar SA, Mathew J, Petrov MS (2017) Age- and sex-specific prevalence of diabetes associated with diseases of the exocrine pancreas: a population-based study. Dig Liver Dis 49:540–544
Collins GS, Reitsma JB, Altman DG, Moons KGM (2015) Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med 162:55–63
Acknowledgements
This study was part of the COSMOS program.
Funding
COSMOS is supported, in part, by the Royal Society of New Zealand (Rutherford Discovery Fellowship to Associate Professor Max Petrov).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Guarantor
The scientific guarantor of this publication is Associate Professor Max Petrov, MD, MPH, PhD.
Conflict of interest
The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.
Statistics and biometry
No complex statistical methods were necessary for this paper.
Informed consent
Institutional Review Board approval was not required because the study was a secondary analysis of the literature.
Ethical approval
Written informed consent was not required for this study because it was a secondary analysis of the literature.
Methodology
• retrospective
• diagnostic or prognostic study
• performed at one institution
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
ESM 1
(DOCX 47 kb)
Rights and permissions
About this article
Cite this article
Abunahel, B.M., Pontre, B., Kumar, H. et al. Pancreas image mining: a systematic review of radiomics. Eur Radiol 31, 3447–3467 (2021). https://doi.org/10.1007/s00330-020-07376-6
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00330-020-07376-6