The purpose to use imaging for staging is to define and classify pancreatic cancer as locally resectable, locally unresectable, and unresectable for distant metastases [11]. The classifications used are the Union Internationale Contre le Cancer, the American Joint Committee Classification, and the Japan Pancreas Society (Pancreatic Cancer Registration Committee of the Japan Pancreas Society 2003). Borderline resectable lesions may undergo resection after chemotherapy and/or radiotherapy [5].

Local staging (T) criteria after recent changes, only consider unresectable tumors those invading either the celiac axis or the superior mesenteric artery. Limited infiltration of venous superior mesenteric vein is now considered resectable using venous grafts [14]. The most useful imaging criteria of vascular invasion is the presence of circumferential involvement >180° [8, 11]. CT and MRI using vascular techniques show sensitivities over 90 % to determine vascular invasion and resectability [15] (Fig. 44.1a, b).

Nodal staging (N) is more challenging for cross-sectional imaging. Size criteria (transverse diameter) show little sensitivity, being a nonspecific criterion [16]. Metastasis in peripancreatic nodes is not important as they will be removed at surgery, being more important to define affectation of celiac node, common artery hepatic node, and para-aortic nodes, as their involvement is a criteria of unresectability [11] (Fig. 44.1b).

Up to 60 % of pancreatic carcinoma presents metastases at diagnosis [8]. The most common distant metastases of pancreatic adenocarcinoma occur to the liver and peritoneum. Therefore, the use of a multiphasic protocol including a portal venous phase allows their assessment in the same study than locoregional staging. MRI is slightly superior to CT in the assessment of liver metastasis using T2-weighted and dynamic contrast-enhanced (DCE)-MRI. In addition, it is well known that the use of DWI and hepatospecific contrast agents improves the detection of liver metastasis, especially if they are used combined [17]. Moreover, DWI has shown higher sensitivity and specificity than CT in a statistically significant manner in the detection of liver metastasis of pancreatic carcinoma [18]. In addition, in a recent series comparing MRI with gadoxetic acid and CT in the assessment of liver metastasis of pancreatic carcinoma, MRI had greater sensitivity for two of the three readers in the MRI-versus-CT analysis (85 % vs 69 %) and for all three readers in the lesion-by-lesion analysis (92–94 % vs 74–76 %) [19].

In the evaluation of peritoneal disease, MRI is considered the best imaging technique, although it is limited in its evaluation (Fig. 44.1c–e). Staging laparoscopy before surgery in resectable disease by preoperative imaging is proposed by some groups [20]. However, a meta-analysis demonstrated that laparoscopy only alters patient management in 4–15 % of the cases after staging CT [21].

The triple-phase enhanced ¹⁸FDG PET-CT is the state-of-the-art protocol for assessment of pancreatic lesions [5]. Pancreatic adenocarcinoma usually shows intense ¹⁸FDG uptake (Fig. 44.2). Although ¹⁸FDG PET-CT has shown its usefulness in the differentiation of chronic and autoimmune pancreatitis (AIP) from pancreatic cancer [22, 23], it has shown a pooled sensitivity and specificity of 90.1 and 80.1 % in the diagnosis of pancreatic cancer [24]. Thus, this functional technique has a limited role in clinical practice for tumor detection and characterization. Limitations of ¹⁸FDG PET-CT are related to false-positive, such as inflammatory tissue, and false-negative results, such as mucinous and necrotic tumors or small tumors located near areas of physiological uptake (ampullary tumors) [25] (Fig. 44.3). The main use of ¹⁸FDG PET-CT is M-staging of pancreatic cancer, as its results for T- and N-staging are far from good [11, 26]. Reported sensitivities and specificities of ¹⁸FDG PET-CT for nodal staging range from 46 to 71 % and from 63 to 100 %, respectively [8]. However, reported sensitivities of 68–70 % for liver metastasis detection (Fig. 44.4) with specificities over 90 % and rates of 25 % in the detection of occult peritoneal disease (Fig. 44.5) make ¹⁸FDG PET-CT suitable for the detection of distant metastases [27, 28]. Furthermore, in a recent series, 12 of 44 patients upstaged from curative to palliative disease after ¹⁸FDG PET-CT [29]. Similarly, in another series, ¹⁸FDG -PET in addition to CT altered the surgical management in 13 % of patients by identifying unsuspected distant metastases or by clarifying the benign nature of indeterminate lesions in CT [30]. Limitations of this technique in the evaluation of liver and peritoneal metastases are related to lesions smaller than 1 cm in the liver and microscopic disease in the peritoneum, due to limitations in spatial resolution.

Another area of interest for ¹⁸FDG PET-CT is the definition of tumor volume and margins in radiotherapy planning. The scarce data available suggest a better performance of ¹⁸FDG PET-CT compared to CT in this task [5]. For example, Topkan and colleagues reported a statistically significant increase in gross tumor volume delineation in 35.7 % of patients with locally advanced pancreatic cancer during radiotherapy planning, in the comparison between CT versus co-registered ¹⁸FDG PET-CT [31]. Interestingly, ¹⁸FDG PET-CT is useful in the early evaluation of treatment response to chemotherapy and radiotherapy, being superior to CT [32, 33]. In addition, in locally advanced unresectable pancreatic cancer, pretreatment SUV values are prognostic of overall and progression-free survival and predictive of histologic response after treatment [5, 34]. Furthermore, initial data supports an independent role for relative change in maximum standardized uptake value (SUV_max) in predicting clinical outcomes in patients with locally advanced pancreatic carcinoma after concurrent chemoradiotherapy [29]. In the depiction of recurrent cancer after treatment, ¹⁸FDG PET-CT is of interest in patients with elevated concentrations of CA 19–9 and those with inconclusive CT findings [35].

In the last years, DWI has expanded its applications outside the brain because of technological improvements in gradient strengths and sequences. Nowadays, DWI forms part of state-of-the-art body MRI protocols. The image contrast on DWI relies on intrinsic differences in the water diffusion between tissues. Furthermore, the role of DWI as a cancer biomarker of tissue cellularity and cell membrane integrity is well recognized [36]. DWI can also help in the diagnosis of pancreatic carcinoma, typically showing high signal intensity with high b values and lower apparent diffusion coefficient (ADC) values than normal pancreas [37, 38] (Fig. 44.6a–c). Lipophilic cell membranes in pancreatic ductal adenocarcinoma, typically a tumor with hypercellular tissue, serve as barriers to free diffusion in the intracellular and extracellular spaces. However, variations in the ADC values of pancreatic adenocarcinoma have been related according to the histopathological composition of the tumors [38, 39]. Normally, dense fibrosis and increased cellular elements make pancreatic cancer to show lower ADC values than the normal pancreatic parenchyma; however, tumors with high content in edematous fibrosis and loose collagen fibers can show higher ADC values than normal pancreas (Fig. 44.3b, c). A recent series demonstrated lack of relationship between ADC values of pancreatic adenocarcinoma and tumor grade or other adverse pathological features [40]. However, the series by Wang and coworkers demonstrated significant lower ADC values for poorly differentiated adenocarcinoma with histopathological characteristics of limited glandular formation and dense fibrosis compared to those of well-/moderately differentiated adenocarcinomas characterized by neoplastic tubular structures [41]. A recent meta-analysis demonstrated sensitivity of 85 % and specificity of 91 % for pancreatic adenocarcinoma detection using DWI with a high b value [42]. The reported sensitivity was similar to that of ¹⁸FDG PET-CT (87 %) with better specificity (83 %). However, the presence of associated acute pancreatitis to pancreatic carcinoma, which more commonly occur in tumors located in the head of the pancreas, may hamper the detection of pancreatic adenocarcinomas at DWI in 47 % of the cases, according to recent data [39]. This is related to increased signal intensity and low ADC value of pancreatic parenchyma distal to pancreatic cancer, demonstrating acute inflammatory changes. At this presentation, pancreatic carcinoma can show ill-defined borders or even low signal intensity at high b value and similar ADC values compared to the inflamed distal pancreatic parenchyma.

Intravoxel incoherent motion (IVIM) model of diffusion signal decay quantitatively assesses the microscopic translational motions that occur in each image voxel. Pure molecular diffusion and microcirculation (also known as blood perfusion) can be distinguished by means of IVIM, using multiple lower and higher b values than 100 s/mm² [43]. In the last years, this model has been proposed for improved pancreatic cancer assessment compared to monoexponential signal decay analysis. Lemke and coworkers demonstrated that perfusion fraction (f) showed more than a tenfold higher significance level in the differentiation between healthy pancreas and pancreatic carcinoma compared with the ADC values [44]. However, another IVIM-derived parameter, tissue diffusivity (D) was unable to differentiate normal tissue and cancer, while ADC could distinguish both entities in a significant manner (Fig. 44.6d). Moreover, the same group of authors has shown that D value rises from moderate to severe fibrosis associated to pancreatic cancer and focal chronic pancreatitis, contrarily to what has been previously shown for ADC values [45].

DWI is also able to accurately differentiate pancreatic carcinoma from benign lesions. A recent meta-analysis reported pooled sensitivity of 86 % and pooled specificity of 91 %, with an area under the curve of the summary receiver operating characteristic of 0.94 [46]. Limitations in the differentiation between mass-forming pancreatitis from poorly differentiated adenocarcinomas using DWI were also a conclusion of this meta-analysis, as both entities show high fibrotic content and low ADC values. In addition, contradictory data in the relation between ADC values of pancreatic cancer and mass-forming pancreatitis have been reported, with an evident overlap in several series [47, 48]. The referred meta-analysis considered as inconsistent the data reported by Huang and colleagues, as in their series DWI was useful in differentiating between pancreatic carcinoma and mass-forming pancreatitis using a high b value of 1,000 s/mm² in combination with ADC quantification in a 3-T magnet [49] (Fig. 44.7a–c). IVIM model has been also tested in this differentiation. In the series by Klauss and colleagues, perfusion fraction was the best parameter in the differentiation between mass-forming pancreatitis and pancreatic carcinoma, while D value was not significantly different between both groups (Figs. 44.6d and 44.7d). Interestingly, ADC calculated with b values between 50 and 300 s/mm² showed statistically significant differences between both entities, which were not possible with ADC calculated with b values of 25, 400, 600, and 800 s/mm² or including all b values. Therefore, the differences between mass-forming pancreatitis and pancreatic cancer are probably due to differences in perfusion [50].

AIP represents a distinct form of chronic pancreatitis which often presents as a pancreatic mass causing obstructive jaundice as well as pancreatic exocrine and endocrine insufficiency. It shows high signal intensity on high b-value DWI sequences and lower ADC values than chronic pancreatitis and pancreatic carcinoma [51–53] (Fig. 44.8). DWI is also useful to monitoring the effect of steroids, as if therapy is successful high-intensity areas on DWI disappear or are markedly decreased, and the ADC values increased almost to the values of normal pancreas [51, 53].

DWI has been proposed for therapy monitoring and prediction of treatment response in different tumors and organ sites [36] (Figs. 44.9 and 44.10). Normally, a positive response to treatment is related to decrease signal on high b value and increase in ADC values, although ADC decreases may occur due to fibrosis formation or necrosis clearance [54]. Preliminary data suggests that a lower ADC calculated using a high b value (1,000 s/mm²) is correlated with early progression in patients with advanced pancreatic cancer treated with chemotherapy [55]. Unfortunately, these data correspond to a unique series including 63 patients, and therefore, further research is needed to validate these results.

The vascular characteristics of pancreatic cancer have been investigated with different imaging methods. Recent advances in technology and contrast agents have made possible to study whole organ perfusion with contrast-enhanced ultrasound (CEUS), CT perfusion, and high temporal resolution DCE-MRI. It is possible to investigate differences in time-intensity curve (TIC) of normal pancreas and focal lesions. Furthermore, semiquantitative or quantitative vascular parameters derived from these techniques have been advocated for tumor characterization, therapy monitoring, and prediction of response to treatment. Technical aspects and clinical applications have been extensively analyzed in previous chapters of this book.

Pancreatic carcinoma is a hypovascular tumor, demonstrating high permeability and limited blood flow. In this situation, volume transfer coefficient (K ^trans) and indirectly, initial area under the concentration curve in 60 s (AUC₆₀), represent blood flow and values of these parameters of pancreatic carcinoma are low (Fig. 44.11). These vascular characteristics have been well demonstrated either by CT perfusion or DCE-MRI [56–58]. In addition, the TIC of pancreatic cancer is quite different to that of mass-forming chronic pancreatitis and neuroendocrine tumors [58, 59] (Fig. 44.12). Mass-forming pancreatitis shows low wash-in with at least a delayed washout, and neuroendocrine tumors show early and fast enhancement and subsequently marked washout. Differently, pancreatic adenocarcinoma demonstrates low wash-in with progressive enhancement or delayed plateau, without washout. Similarly, CEUS has shown to increase the diagnostic confidence in the differentiation between pancreatic carcinoma and mass-forming pancreatitis [60]. Furthermore, vascular parameters derived from CT perfusion or DCE-MRI help in the differentiation of pancreatic solid tumors. Therefore, it may be concluded that perfusion studies are useful in the differentiation of these entities, although conventional CT and MRI protocols are also able to perform accurate characterization (Fig. 44.13). Moreover, recently vascular parameters derived from CT perfusion were able to significantly distinguish between low- and high-grade pancreatic adenocarcinoma [61]. Similar results have been found with CEUS, which show a positive correlation between the type of vascularization in pancreatic adenocarcinoma with histology grade and mean vessel density [62]. In addition, CT perfusion parametric maps help in the detection of pancreatic carcinomas which are not visible on multiphasic CT [63].

Initial data have related high pretreatment K ^trans values of locally advanced pancreatic carcinoma to good response to treatment, as these elevated values of K ^trans could represent better delivery of the drugs to the tumor [64, 65]. Park and colleagues evaluated 30 patients with locally advanced pancreatic carcinoma with CT perfusion before and 3 months after gemcitabine-based concurrent chemotherapy and radiation therapy. Responders demonstrated higher tumor K ^trans and blood volume values, although there was only a significant difference for K ^trans [64]. Similarly, Akisik and coworkers evaluated 11 patients with locally invasive pancreatic cancer before and 28 days after combined chemotherapy and antiangiogenic therapy with DCE-MRI analyzed using a two-compartment kinetic model. Pretreatment K ^trans and K ^ep were significantly higher in tumors that showed good response than in nonresponders. An elevated pretreatment K ^trans was 100 % sensitive and 71 % specific to predict tumor response. Furthermore, in responders, a significant decrease of K ^trans, v _e, C _peak, slope, and AUC₆₀ was reported [65].