key: cord-0028899-g8pk65uq authors: Casà, Calogero; Piras, Antonio; D’Aviero, Andrea; Preziosi, Francesco; Mariani, Silvia; Cusumano, Davide; Romano, Angela; Boskoski, Ivo; Lenkowicz, Jacopo; Dinapoli, Nicola; Cellini, Francesco; Gambacorta, Maria Antonietta; Valentini, Vincenzo; Mattiucci, Gian Carlo; Boldrini, Luca title: The impact of radiomics in diagnosis and staging of pancreatic cancer date: 2022-03-16 journal: Ther Adv Gastrointest Endosc DOI: 10.1177/26317745221081596 sha: a86136edf5a9a925b7381eca336e39b615c00fdf doc_id: 28899 cord_uid: g8pk65uq INTRODUCTION: Pancreatic cancer (PC) is one of the most aggressive tumours, and better risk stratification among patients is required to provide tailored treatment. The meaning of radiomics and texture analysis as predictive techniques are not already systematically assessed. The aim of this study is to assess the role of radiomics in PC. METHODS: A PubMed/MEDLINE and Embase systematic review was conducted to assess the role of radiomics in PC. The search strategy was ‘radiomics [All Fields] AND (“pancreas” [MeSH Terms] OR “pancreas” [All Fields] OR “pancreatic” [All Fields])’ and only original articles referred to PC in humans in the English language were considered. RESULTS: A total of 123 studies and 183 studies were obtained using the mentioned search strategy on PubMed and Embase, respectively. After the complete selection process, a total of 56 papers were considered eligible for the analysis of the results. Radiomics methods were applied in PC for assessment technical feasibility and reproducibility aspects analysis, risk stratification, biologic or genomic status prediction and treatment response prediction. DISCUSSION: Radiomics seems to be a promising approach to evaluate PC from diagnosis to treatment response prediction. Further and larger studies are required to confirm the role and allowed to include radiomics parameter in a comprehensive decision support system. Pancreatic cancer (PC) is one of the most aggressive tumours, representing the fourth cause of cancer-related deaths with 132,600 estimated new diagnoses, 128,000 deaths per year and accounting for 3.4-6.6% of all cancer cases, respectively. 1, 2 The incidence and mortality are slightly increasing during the years with a 5-year overall survival (OS) rate lower than 10%. 3 Males are more frequently affected than females and incidence in both sexes increases with age, reaching its peak in people older than 70 years. 4 Primary prevention is of utmost importance as there is currently no effective method of screening for the general population. Nevertheless, molecular or imaging tests can help early diagnosis in high-risk cohorts; for example, endoscopic ultrasound (EUS) and magnetic resonance imaging (MRI) allowed to detect precursor or invasive pancreatic neoplasms in asymptomatic people with an inherited predisposition (5-10% of all patients), thus increasing the resectability of the tumours and the overall prognosis of these patients. 5 Further data are needed to establish the most adequate approach and to select patients who could benefit the most from screening programmes. Signs and symptoms are indeed not disease-specific and usually appear when it has already become unresectable or distantly spread. Computed tomography (CT) is the most commonly used imaging diagnostic tool, supported The impact of radiomics in diagnosis and staging of pancreatic cancer 2 journals.sagepub.com/home/cmg by EUS with fine needle biopsy or aspiration in the identification of small lesions and in providing a definitive diagnosis. 6 MRI and positron emission tomography (PET) may also contribute to the systemic staging of the disease and in characterizing the primary tumour as resectable, borderline resectable or unresectable. PC is a heterogeneous group of diseases; nearly 90% of the total are adenocarcinomas. Exocrine tumours are known to have a more aggressive behaviour compared with endocrine ones, with a median survival of about 4 versus 27 months, as reported in a large population-based study. 7 Pancreatic adenocarcinoma can progress from different types of precursor lesions, such as pancreatic intraepithelial neoplasia (PanIN) and intraductal papillary mucinous neoplasia (IPMN), under the push of consecutive genetic alterations whose significance is frequently unknown. 8 Some of these mutations have been recognized as having a role in different steps of carcinogenesis or even a prognostic significance, such as SMAD4 loss, which is linked to metastatic spread and lower survival rates. 9 These alterations may have, therefore, the potential to identify precursor lesions and serve as targets for new therapies. Currently, the first treatment option for PC is represented by radical surgery. Unfortunately, few more than 10% of patients are amenable to curative-intent surgery and prognosis remains poor even in this cluster of patients, with 5-year survival rates around 20%. 10 The quantitative analysis of medical images data and the extraction of imaging features, also called 'radiomics', represent an emerging approach in personalized medicine and advanced diagnostics, especially for disease characterization or outcome prediction. [11] [12] [13] The interest towards radiomics is rapidly growing in the multidisciplinary cancer community as it shows an interesting pertinency and efficacy to answer several clinical questions arising in the management of patients affected by other gastrointestinal tumours. [14] [15] [16] [17] [18] [19] [20] [21] The aim of this study is to systematically assess and summarize evidence published in the scientific literature about the different applications of radiomics in PC. We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) recommendations. 22 All the papers were selected and analyzed by a board of five radiation oncologists (ROs; CC, AD, FP, SM, AP), followed by an independent validation of three experts in PC (two ROs, GCM and FC, and one gastroenterologist, IB) and three experts on radiomics (two ROs, ND and LB, for the clinical point of view, and one physicist, DC, for the technical point of view). The whole process, the results and the discussion about potential discrepancies were validated by the other two different independent expert ROs (MAG and VV). A total of 123 studies were obtained using the mentioned search strategy on PubMed/ MEDLINE, and 183 articles were obtained on Embase. Of these, 68 papers were selected based on the title according to the previously described criteria. The selection process, shown in Figure 1 , led to the identification of 58 potentially eligible studies according to the abstract (two papers discarded according to the exclusion criterion 1; seven papers discarded for the exclusion criterion 2, and one paper discarded according to the exclusion criterion 3). After full-text analysis, one paper was discarded for the exclusion criterion 1 and one paper according to criterion 2. After the complete selection process, a total of 56 papers were considered eligible for the analysis of the results. All the studies included in the analysis are retrospective; the range of publication year is from 2017 to 2021. The participant centres' nations were China (n = 26; 46.4%), the United States (n = 13; 23.2%), Italy (n = 6; journals.sagepub.com/home/cmg 3 10.7%), Canada and Germany (n = 3; 5.4%), Japan (n = 2; 3.6%), France, Russia and South Korea (n = 1; 1.78%). The radiomics analysis was performed on CT (n = 43; 76.8%), PET-CT (n = 6; 10.7%) or MR (n = 7; 12.5%) images. The median number of patients involved in the analysis was 110 (range 10-422); the median number of features analyzed was 410 (range 1-3328). Radiomics, as resulted from the analyzed works, plays a role in many aspects of PC management. Diagnostic imaging features were found to correlate with diagnosis and differentiation of pancreatic ductal adenocarcinoma (PDAC) from noninvasive diseases in 10 studies (17.9%). [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] CT-based features were found able to distinguish PC from normal pancreatic tissue 23, 28 and to detect cancer in the context of intraductal papillary mucinous neoplasm (IPMN) 27, 30 ; CT-and PET-CTbased radiomics could discriminate PC from pancreatitis, [24] [25] [26] 31, 32 while MRI was able to discriminate pancreatic neuroendocrine tumours (pNETs) from solid pseudopapillary tumours in one study. 29 Six studies (10.7%) analyzed technical feasibility and reproducibility aspects of radiomics analysis. Yamashita et al. 33 demonstrated that the variation between contrastenhanced CT scans (e.g. scanner model, pixel spacing and contrast administration rate) affects radiomic feature reproducibility more significantly than the variation in segmentation. Mori Other two studies proposed PET-CT-or CT-based radiomics to predict mutational status or PD-L1 expression. 55 In total, 21 studies (37.5%) described the application of radiomics in PC risk group stratifications, mainly in preoperative setting. 40,47,48, [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] Within this heterogeneous group of papers, four articles 62, 66, 72, 74 analyzed the role of CT-based radiomics in lymph nodes (LNs) metastasis risk assessment before surgery. The preoperative evaluation of the risk of a positive resection margin using CT-based radiomics analysis was the aim of three papers. 58 Table 1 summarizes the main characteristics of the analyzed studies, including aim, conclusion and the cluster area of radiomic correlation. Radiomics seems to be an effective approach to evaluate patients affected by PC in several clinical settings and in different clinical contexts, from diagnosis to risk stratification, from biologic or genomic status prediction to treatment response evaluation or assessment. Figure 2 shows an example of clinical implementation of a delta radiomic model during a stereotactic radiotherapy treatment prescribed in five fractions. The common lack of technical standardization of the features extraction process and the necessity of external independent validation for each proposed predictive model hamper the feasibility of extensive radiomics studies in PC and impose the involvement of multidisciplinary research teams and the enrolment of larger patient samples. Another limitation that reduces the generalizability of these observations is represented by the multiple kinds of image modalities and, for each of them, the variability of acquisition protocols. Moreover, the huge heterogeneity of features analyzed in the different studies poses a limitation in the possibility of creating a cluster of significant features for radiomics in this pathology. The same number of features analyzed, study by study, is very variable (from 1 to 3328 features per study) documenting that the resulting models are very much linked to the analysis experience of each centre. Despite the above limitations, radiomics could potentially have an important role in providing reliable risk stratification (for both outcomes or complications), facilitating surgical choices, predicting clinical response after treatments, allowing differential diagnosis between cancer and other benign pancreatic abnormalities and predicting histological examination, disease differentiation grade or specific gene mutations. The multidisciplinary management of these patients by dedicated teams of different specialists The nomogram achieved an optimal preoperative, noninvasive differential diagnosis between atypical pNET and PDAC. Gu et al. The portal rad-score is significantly associated with the pathologic SMV resection margin, and it can accurately and noninvasively predict the SMV resection margin in patients with PC. (Continued) (Continued) Example of clinical implementation of a delta radiomic model during a stereotactic radiotherapy treatment prescribed in five fractions. Before the start of the treatment, a radiomic model able to predict the LC 1 year from the end to the treatment was trained and tested on a retrospective cohort of patients. The model was based on the radiomic analysis of the MR images acquired during simulation and during fractions 1 and 2. Using the radiomic model, the RO can have a prediction of 1 year LC at the end of the fraction 2, so having the possibility to modify the radiation treatment for the remaining three fractions, increasing the dose or moving towards alternative approaches. journals.sagepub.com/home/cmg 15 has proven to be a promising approach in terms of treatment quality and outcomes. 79 Radiation therapy (RT) can play a role in all possible scenarios, being particularly relevant in reducing the risk of relapse of resected cancer (adjuvant setting) and in pursuing LC in locally advanced disease (definitive setting). Chemotherapy is also part of the management in all stages, and is administered in association with other treatment modalities (neoadjuvant, adjuvant therapy) or as a main part of the treatment for metastatic disease. An important role to improve diagnostic and therapeutic options for PC patients is represented by the innovative approach of personalization through risk stratification, aiming to better define patients risk category and choose therapeutic patterns accordingly. 80, 81 Patients' stratification is mainly performed based on the genotype, but the lack of a convergency in genomics models seems to reveal that the understanding of biological and clinical heterogeneity of PC is still far from completely understood. 82, 83 Recently, advances in imaging with the possibility of combining endoscopic modalities with image fusion techniques have provided new ways of approaching the diagnosis and stratification of PC. [84] [85] [86] As part of an omics-guided care pathway, the integration of these modalities with artificial intelligence and deep-learning methods would seem to be a compelling prospect for pursuing personalized care in a highly heterogeneous disease. [87] [88] [89] In this framework, radiomics is a promising method to be investigated that could provide more information about PC patients and its future integration in multiomics clinical support systems will allow more personalized and efficacious cancer care. Cancer incidence and mortality patterns in Europe: estimates for 40 countries and 25 major cancers in 2018 Cancer statistics, 2020 Pancreatic Cancer: Cancer Stat facts Epidemiology of pancreatic cancer Pancreatic cancer screening Imaging of pancreatic cancer: an overview Differences in survival by histologic type of pancreatic cancer SMAD4 and its role in pancreatic cancer The impact of curative intent surgery on the survival of pancreatic cancer patients: a U.S. population-based study Radiomics: images are more than pictures, they are data Moddicom: a complete and easily accessible library for prognostic evaluations relying on image features The image biomarker standardization initiative: standardized quantitative radiomics for highthroughput image-based phenotyping Magnetic resonance, vendor-independent, intensity histogram analysis predicting pathologic complete response after radiochemotherapy of rectal cancer Radiomics for rectal cancer Towards a modular decision support system for radiomics: a case study on rectal cancer Fractal-based radiomic approach to predict complete pathological response after chemoradiotherapy in rectal cancer Delta radiomics for rectal cancer response prediction with hybrid 0.35 T magnetic resonance-guided radiotherapy (MRgRT): a hypothesis-generating study for an innovative personalized medicine approach CT and MRI of pancreatic tumors: an update in the era of radiomics Pancreas image mining: a systematic review of radiomics Delta radiomics can predict distant metastasis in locally advanced rectal cancer: the challenge to personalize the cure Preferred reporting items for systematic reviews and metaanalyses: the PRISMA statement Utility of CT radiomics features in differentiation of pancreatic ductal adenocarcinoma from normal pancreatic tissue Radiomics analysis for the differentiation of autoimmune pancreatitis and pancreatic ductal adenocarcinoma in 18F-FDG PET/CT Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features Differentiation of focal-type autoimmune pancreatitis from pancreatic ductal adenocarcinoma using radiomics based on multiphasic computed tomography Multiphase computed tomography radiomics of pancreatic intraductal papillary mucinous neoplasms to predict malignancy A novel multiresolution-statistical texture analysis architecture: radiomics-aided diagnosis of PDAC based on plain CT images Radiomics analysis based on diffusion kurtosis imaging and T2 weighted imaging for differentiation of pancreatic neuroendocrine tumors from solid pseudopapillary tumors A combined nomogram model to preoperatively predict histologic grade in pancreatic neuroendocrine tumors CT radiogenomic characterization of the alternative lengthening of telomeres phenotype in pancreatic neuroendocrine tumors CT-radiomic approach to predict G1/2 nonfunctional pancreatic neuroendocrine tumor Imaging phenotype using 18F-fluorodeoxyglucose positron emission tomography-based radiomics and genetic alterations of pancreatic ductal adenocarcinoma Radiogenomics for predicting p53 status, PD-L1 expression, and prognosis with machine learning in pancreatic cancer CT-based radiomics score for distinguishing between grade 1 and grade 2 nonfunctioning pancreatic neuroendocrine tumors Magnetic resonance imaging radiomic analysis can preoperatively predict G1 and G2/3 grades in patients with NF-pNETs Noncontrast radiomics approach for predicting grades of nonfunctional pancreatic neuroendocrine tumors Development and multicenter validation of a CT-based radiomics signature for discriminating histological grades of pancreatic ductal adenocarcinoma A radiomicsbased formula for the preoperative prediction of postoperative pancreatic fistula in patients with pancreaticoduodenectomy Contrast-enhanced CT radiomics for predicting lymph node metastasis in pancreatic ductal adenocarcinoma: a pilot study Development of a novel multiparametric mri radiomic nomogram for preoperative evaluation of early recurrence in resectable pancreatic cancer Pancreatic ductal adenocarcinoma: a radiomics nomogram outperforms clinical model and TNM staging for survival estimation after curative resection CNN-based survival model for pancreatic ductal adenocarcinoma in medical imaging Relationship between radiomics and risk of lymph node metastasis in pancreatic ductal adenocarcinoma Performance of CT-based radiomics in diagnosis of superior mesenteric vein resection margin in patients with pancreatic head cancer Prognostic value of CT radiomic features in resectable pancreatic ductal adenocarcinoma Identification of pancreaticoduodenectomy resection for pancreatic head adenocarcinoma: a preliminary study of radiomics Dual tracer 68Ga-DOTATOC and 18F-FDG PET/ computed tomography radiomics in pancreatic neuroendocrine neoplasms: an endearing tool for preoperative risk assessment Training and validation of a robust PET radiomic-based index to predict distant-relapse-free-survival after radio-chemotherapy for locally advanced pancreatic cancer A Radiomics nomogram for the preoperative prediction of lymph node metastasis in pancreatic ductal adenocarcinoma Prognostic value of FDG-PET radiomics with machine learning in pancreatic cancer Applying a radiomicsbased strategy to preoperatively predict lymph node metastasis in the resectable pancreatic ductal adenocarcinoma Validation of prognostic radiomic features from resectable pancreatic ductal adenocarcinoma in patients with advanced disease undergoing chemotherapy Radiomics-assisted presurgical prediction for surgical portal veinsuperior mesenteric vein invasion in pancreatic ductal adenocarcinoma Predictive modeling for voxel-based quantification of imaging-based subtypes of pancreatic ductal adenocarcinoma (PDAC): a multi-institutional study Radiomics facilitates candidate selection for irradiation stents among patients with unresectable pancreatic cancer Cassadie Moravek ¥ pancreatic cancer action network continue NCCN guidelines version 1 Genomic analyses identify molecular subtypes of pancreatic cancer Subtypes of pancreatic ductal adenocarcinoma and their differing responses to therapy Pancreatic cancer biology and genetics from an evolutionary perspective Intratumor heterogeneity and branched evolution revealed by multiregion sequencing How to get away with COVID-19: endoscopy during post-peak pandemic CT and MRI slice separation evaluation by LabView developed software A guide for effective anatomical vascularization studies: useful ex vivo methods for both CT and MRI imaging before dissection Deep learning: a review for the radiation oncologist Artificial intelligence (AI) and interventional radiotherapy (brachytherapy): state of art and future perspectives Recommendations on how to establish evidence from auto-segmentation software in radiotherapy Visit SAGE journals online journals.sagepub.com/ home/cmg The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.