Transabdominal US represents the first diagnostic approach in most cases. It is a non-invasive, radiation-free, readily available and inexpensive imaging technique enabling the study the entire abdomen, including pancreatic lesions. However, it is considered extremely dependent on operator expertise and limited by patient characteristics such as bowel gas and obesity (for instance, most patients with insulinoma tend to gain weight, and this limits the value of US [4]).

Usually pancreatic NETs (P-NETs) appear as a homogenous hypoechoic nodular lesion with regular margins. Larger lesions may appear with a more heterogeneous echogenicity, probably due to different internal composition such as greater amounts of hyalinized stroma, cystic degeneration or bleeding due to the fragility of new blood vessels; any cystic areas within the lesion will have a hypo-anechoic appearance [5]. After the administration of contrast medium, during contrast-enhanced US (CEUS) examination, P-NETs show a hypervascular behavior, enhancing avidly in comparison with the surrounding parenchyma. Particular attention should be paid to enlarged and rounded locoregional lymph nodes and to the presence of liver metastases, considered indirect signs of malignant transformation.

Transabdominal US has a sensitivity ranging between 20% and 80% for the detection of P-NETs [6, 7]. This wide range could be explained by the great sensitivity in detecting rare large histotypes (85–95%) and intermediate-to-low sensitivity for the most common histotypes (around 50%), with a variable range of 19–40% for gastrinomas and 25–64% for insulinomas [8].

MDCT is the first-line non-functional imaging technique when a P-NET is clinically suspected. The CT protocol generally includes an unenhanced phase, a delayed arterial phase (specific to assess pancreatic parenchyma, about 40 s after i.v. injection of contrast medium), a portal venous phase (55–70 s after i.v. injection of contrast medium) and a delayed phase (120 s after i.v. injection of contrast medium) [9, 10].

Functioning P-NETs during the unenhanced phase are usually isodense to the parenchyma and this can be useful to detect hemorrhage or calcifications. The arterial phase is considered the most sensitive acquisition phase. In this phase, the hypervascular feature of P-NETs makes even a rounded smooth lesion hyperattenuating with the best attenuation difference compared to the surrounding pancreatic parenchyma (Fig. 7.1). During the venous phase the attenuation of both the tumor and the normal parenchyma generally decreases, so that the lesion is less detectable [8]. However, in larger tumors (e.g., glucagonomas) it is possible to observe increased tissue degeneration and consequent heterogeneity of density during both phases [1].

Moreover the portal venous phase can be, in a minority of cases, more useful for identifying liver metastases, evaluating locoregional lymphadenopathy and assessing the relationship between the tumor and the surrounding structures. For this reason, biphasic imaging after i.v. injection of contrast medium is currently recommended to improve detection and characterization [11].

Regarding non-functioning tumors, they are larger in size (average diameter, 4 cm) at the time of detection, due to the absence of symptoms, and well defined, encapsulated and with heterogeneous enhancement. Heterogeneity can be related to degenerative cystic areas, necrotic regions and, rarely, fibrosis. Even in the remote case that they are completely cystic, a hypervascular rim is detectable in up to 90% of cases. Moreover, tumor local aggression is evincible through retroperitoneal invasion or metastases (up to 80% of cases) to regional lymph nodes or to the liver. It is very uncommon to find pancreatic or biliary duct obstruction [1].

In recent studies, the diagnostic accuracy and sensitivity of MDCT in the assessment of P-NETs has improved from a range of 14–30% (reported in older studies) to 69–94% [1]. Moreover, the wide range in sensitivity can also be interpreted observing that many small P-NETs are missed on CT [12]. Recently, sensitivity has improved considerably. In a recent study comparing MDCT and dual-energy CT, MDCT had a sensitivity of 68.8%, while dual-energy CT reached 95.7% [13]. An important limitation to this technique is radiation exposure, in particular for repeated follow-up examinations. By applying new dose reduction techniques, such as iterative reconstruction, it would be possible to reach a low dose exposure with a good diagnostic image quality [14, 15].

MRI plays an important role in the characterization of P-NETs. Indications include localization of suspected lesions not clearly characterized by US or CT and avoidance of radiation dose in young patients, especially during follow-up [8].

An appropriate protocol for the study of pancreatic parenchyma and specifically of P-NETs includes T1- and T2-weighted sequences, both with and without fat suppression, diffusion-weighted imaging (DWI) and postcontrast T1-weighted fat-suppressed dynamic imaging after the injection of gadolinium contrast medium. In the assessment of P-NETs, unenhanced T1-weighted fat-suppressed MRI has marked sensitivity, reaching 75% [6]. P-NETs usually show a decreased signal intensity on T1-w sequences and high signal intensity on T2-w sequences relatively to the normal pancreatic gland; after i.v. injection of gadolinium-based contrast media they usually show homogeneous enhancement. However, due to the wide histological heterogeneity of P-NETs, the signal intensity and enhancement patterns might be varied: cystic lesions may show a thin rim enhancement and appear bright on T2-w images, tumors rich in collagen or fibrous tissue could present low signal on T2-w images (Fig. 7.2). MRI, because of the higher soft tissue contrast compared with US and CT, usually allows a better evaluation in the case of multiple NET lesions [16]. Regarding sensitivity and specificity, MRI has improved its accuracy reaching a range of 74–94% and 78–100%, respectively [1].

A relatively new tool to be implemented in all MRI acquisition protocols, is DWI. This is a novel bioimaging technique that allows one to evaluate quantitatively the hypercellularity of neoplastic lesions. DWI has a diagnostic role in the case of negative or doubtful imaging findings to support the diagnostic process especially in non-hypervascular lesions [8].

Moreover, a cholangiopancreatography (MRCP) sequence should be included in the protocol to assess the relation between the tumor and the pancreatic and main bile ducts [16].

Endoscopic US (EUS) has been increasingly used in the localization of P-NETs, particularly for the detection of small insulinomas [8]. Furthermore, EUS plays an important role in early detection and follow-up of multifocal lesions that are common among patients with MEN1 and von Hippel-Lindau syndromes [17]. The close proximity of the pancreas to the stomach and the duodenum allows a detailed examination of small lesions, especially those located at the pancreatic head. EUS has the best spatial resolution compared to other different imaging modalities: it can detect lesions smaller than 0.2 cm as well as adjacent celiac, peripancreatic, para-aortic, and periportal lymphadenopathies, and it can assess vascular invasion [18]. Typically, small P-NETs appear as rounded, homogeneous, hypoechoic lesions. If calcifications are present, a diagnosis of somatostatinoma can be made.

Rösch et al. [18] in 1992 reported an EUS sensitivity of 82% and specificity of 92% in patients whose tumors had previously remained undetected by other different methods. More recent studies reported a diagnostic accuracy ranging from 79% to 94% [19–22].

Another very important advantage of EUS is the possibility to perform fine-needle aspiration (FNA) during the procedure. Using a 22- or 25-gauge needle, a tissue sample can be obtained for cytological evaluation and immunohisto-chemistry. With this combined approach, sensitivity and specificity improve up to 84% and 92.5%, respectively [23].

An accurate preoperative vascular assessment prior to pancreatic surgery is of utmost importance. Angiography, performed through selective intra-arterial injection of contrast medium, is considered the best method to investigate vascular anatomy and it improves detection of duodenal and pancreatic gastrinomas [24]. Neuroendocrine tumors are seen on arteriography as diffusely enhancing masses without tumor vessels and without arteriovenous shunting.

Furthermore angiography can lead to several interventional procedures. Preoperative angioembolization of hypervascular tumors was proven safe and may result in a decreased risk of bleeding during surgery. Embolization of primary pancreatic cancer was described by Hirose et al. [25] and Ben-Ishay et al. [26] who reported on preoperative angioembolization of a hypervascular P-NET located at the head of the pancreas. Preoperative angioembolization of the hepatic artery prior to en bloc celiac axis resection for pancreatic body cancer was also reported [27].

Transabdominal US has the great advantage of not using ionizing radiation, and it is extremely useful especially in young patients. On the other hand, its role in gastrointestinal primary NETs is generally limited due to the presence of bowel gas artifacts that usually interfere with an accurate bowel wall evaluation [29]. For those reasons, a very large range of sensitivity (15–80%) in the detection of gastroenteric NETs (GE-NETs) has been reported, not only due to small tumor size, but also due to anatomical location and operator experience [29].

The procedure is quite simple: the probe (3–5 MHz) is placed directly on the abdomen; the bowel loops can be distended before the examination through oral administration of polyethylene glycol [1, 29]. Usually GE-NETs appear as hypoechoic nodules, often with a hyperechoic rim [1, 29]. CEUS improves the sensitivity and specificity by highlighting the hypervascular behavior of these lesions, especially if small. US can also been used during interventional procedures to guide biopsy [1].

GE-NETs are most commonly studied by using CT for tumor localization, staging, and follow-up during and after therapies [29, 30]. MDCT scanners have optimal spatial resolution (<0.6 mm for most modern CT scanners) and allow acquisition of the entire abdomen in a single breath-hold and reconstruction of images on multiple planes (coronal, axial, sagittal), extremely useful to improve lesion detection and to evaluate relationships with adjacent abdominal organs. Moreover, by applying advanced reformatting techniques (curved reformats, three-dimensional volume rendering and maximum intensity projection) it is possible to obtain an accurate evaluation of vascular structures, contributing to plan properly surgical procedures [1, 29, 31].

A specific CT acquisition protocol is crucial to improve diagnostic accuracy. A good evaluation of the small bowel requires a combination of two important conditions: a fast imaging technique and good luminal distension through the administration of enteric contrast agents (water, methylcellulose, solutions containing locust bean gum, mannitol, barium sulfate, and polyethylene glycol). The most accepted filling strategy involves a 6-hour fast prior to the examination and a large volume of enteral contrast material (1500–2000 mL) administered orally during the 40 min before the examination. Spasmolytics are useful for reducing bowel peristalsis and motion artifacts. Furthermore, the use of i.v. injection of contrast medium is mandatory for the assessment of bowel walls, lesion enhancement, and mesenteric vessels [32]. The CT protocol generally includes an unenhanced phase, an arterial phase (the most sensitive phase to assess lesion enhancement, about 20 s after i.v. injection of contrast medium), a portal venous phase (55–70 s after i.v. injection of contrast medium) and a delayed phase (120 s after i.v. injection of contrast medium) [29, 33]. It is important to underline that differences in time delays after i.v. injection of contrast media, especially for the arterial phase, could substantially affect image quality, risking potentially both false-positive or false-negative findings [29].