Role of Endoscopic Ultrasound in the Diagnosis of Pancreatic Neuroendocrine Neoplasms

Although pancreatic neuroendocrine neoplasms (PNENs) are relatively rare tumors, their number is increasing with advances in diagnostic imaging modalities. Even small lesions that are difficult to detect using computed tomography or magnetic resonance imaging can now be detected with endoscopic ultrasound (EUS). Contrast-enhanced EUS is useful, and not only diagnosis but also malignancy detection has become possible by evaluating the vascularity of tumors. Pathological diagnosis using EUS with fine-needle aspiration (EUS-FNA) is useful when diagnostic imaging is difficult. EUS-FNA can also be used to evaluate the grade of malignancy. Pooling the data of the studies that compared the PNENs grading between EUS-FNA samples and surgical specimens showed a concordance rate of 77.5% (κ-statistic = 0.65, 95% confidence interval = 0.59–0.71, p < 0.01). Stratified analysis for small tumor size (2 cm) showed that the concordance rate was 84.5% and the kappa correlation index was 0.59 (95% confidence interval = 0.43–0.74, p < 0.01). The evolution of ultrasound imaging technologies such as contrast-enhanced and elastography and the artificial intelligence that analyzes them, the evolution of needles, and genetic analysis, will further develop the diagnosis and treatment of PNENs in the future.


Introduction
Pancreatic neuroendocrine neoplasms (PNENs) are relatively rare tumors that account for 2-3% of all pancreatic tumors. However, the number of reported cases has been increasing, mainly because of the advances in various diagnostic imaging modalities. Among them, endoscopic ultrasound (EUS) has a superior sensitivity for detecting PNENs compared with computed tomography (CT) and magnetic resonance imaging (MRI). With its high resolution, and when performed by experienced hands, EUS can detect focal lesions as small as 2-5 mm [1]. Tissue acquisition using EUS with fine-needle aspiration (EUS-FNA) is essential for the diagnostic and treatment decisions. Here, we review the current literature regarding the role of EUS in the diagnosis of PNENs. Since this study focuses on diagnosis, it does not include interventional EUS, such as EUS-ablation.

Types of EUS
EUS, in which the tip of the endoscope contains a high-frequency transducer, provides high-resolution images of the pancreaticobiliary region. There are two types of scope: radial scanning (RS) and curved linear array (CL). The wide 360 • scanning range of an RS scope makes it easy to grasp relationships with surrounding organs and blood vessels. It is comparatively easier to visualize imaging in affigurelignment with the organ axis [2]. By contrast, the scanning range of the CL scope is narrow (180 • ), and it is difficult to align with the organ axis. However, a study reported the superiority of the overall imaging

EUS for Detecting PNENs
EUS enables detailed observation of the entire pancreas with high tissue resolution, without being affected by the gastrointestinal tract or subcutaneous fat (Figure 1a). In a systematic review, Puli et al. reported that EUS had a sensitivity of 87.2% and a specificity of 98.0% when used for the detection of PNENs [4]. Manta et al. reported that CT failed to detect 68.4% of PNENs < 10 mm and 15% of PNENs ≤ 20 mm in diameter [5]. Moreover, it has been reported that the sensitivity of CT is reduced for small lesions < 1 cm and that 91% of PNENs that are difficult to detect using multidetector-row CT can be detected with EUS [6]. James et al. reported in a meta-analysis that preoperative EUS consistently increased the overall PNEN detection rate by > 25% after a CT scan, with or without additional investigative modalities such as MRI or ultrasound [7]. Thus, EUS is an essential modality for the detection of small PNENs. The role of EUS surveillance has been controversial in recent years. PNENs can occur sporadically or as part of the hereditary multiple tumor syndromes: von Hippel-Lindau (VHL) disease and multiple endocrine neoplasia type 1 (MEN1). MEN1 or VHL patients with a PNEN may undergo imaging of the pancreas every 6 to 12 months to assess the growth rate of the tumor. Concerning the radiation risk in younger patients, EUS and MRI are preferred. In patients with MEN1, the superiority of EUS has been reported for the detection of PNENs [8][9][10][11]. However, Daskalakis et al. reported that MRI is better than EUS for the detection and subsequent surveillance of MEN1-related PNENs larger than 10 mm and it seems to be cost-effective [9]. Kappelle et al. showed that small PNENs in patients with MEN1 grow more slowly than previously suggested. The necessity of EUS The role of EUS surveillance has been controversial in recent years. PNENs can occur sporadically or as part of the hereditary multiple tumor syndromes: von Hippel-Lindau (VHL) disease and multiple endocrine neoplasia type 1 (MEN1). MEN1 or VHL patients with a PNEN may undergo imaging of the pancreas every 6 to 12 months to assess the growth rate of the tumor. Concerning the radiation risk in younger patients, EUS and MRI are preferred. In patients with MEN1, the superiority of EUS has been reported for the detection of PNENs [8][9][10][11]. However, Daskalakis et al. reported that MRI is better than EUS for the detection and subsequent surveillance of MEN1-related PNENs larger than 10 mm and it seems to be cost-effective [9]. Kappelle et al. showed that small PNENs in patients with MEN1 grow more slowly than previously suggested. The necessity of EUS surveillance for MEN1 patients with only small asymptomatic PNENs may be reduced [12].

Precautions for EUS in Functional PNENs
Nonfunctional PNENs are often asymptomatic and undetected until the tumors have grown large enough to cause a mass effect or until they metastasize. However, because functioning PNENs secrete hormones that lead to symptoms, their presence is suspected earlier, and a diagnosis is often made when the lesions are small. Functioning tumors comprise 34.5% of all PNENs [13]. Insulinomas are the most common functional tumors (20.9%), followed by gastrinomas (8.2%), glucagonomas (3.2%), vasoactive intestinal peptide-producing tumors (0.6%), and somatostatinomas (0.3%). Insulinomas tend to be smaller than other functioning PNENs because of the dramatic clinical syndrome caused by insulin secretion. Insulinomas are usually small at diagnosis, with 90% of tumors < 2 cm in diameter and 40% < 1 cm [14]. Most insulinomas occur in the pancreas, and these tumors are evenly distributed throughout the pancreatic head, body, and tail [15]. Approximately 2-10% of patients with insulinomas have multiple tumors, particularly those with MEN1 [14,16]. Thus, it is important to evaluate the entire pancreas in detail when observing insulinomas with EUS. Although insulinomas are usually sporadic, they account for 10-30% of the functioning PNENs in patients with MEN1. MEN1 coexists in 4-5% of insulinomas [15].
Gastrinomas often arise in the gastrinoma triangle, an area bounded by the junctions of the cystic duct and the common bile duct superiorly, the second and third portions of the duodenum inferiorly, and the neck and body of the pancreas medially [14]. They are more common in the duodenum than in the pancreas. Approximately 80% of sporadic lesions and 90% of lesions associated with MEN1 originate from the duodenum. Previously, the pancreas was believed to be the most common location; however, many of these masses may have been peripancreatic nodal metastases from small duodenal tumors. Pancreatic gastrinomas have an average diameter of 3-4 cm, and most are located in the pancreatic head. Duodenal gastrinomas are usually < 1 cm in diameter and are often multicentric, especially in patients with MEN1 [14]. Under EUS, it is necessary to carefully observe not only the pancreatic head but also the duodenal wall. Although most gastrinomas sporadically arise, they are the most common functioning PNENs in patients with MEN1 (20-25% of all gastrinomas occur in these patients).
As mentioned above, because PNENs and MEN1 are related, we must not forget to check for the coexistence of MEN1 in PNEN patients. Examples of multiple NENs associated with MEN-1 from our department are shown in Figure 2.

Role of EUS Elastography
Elastography measures the stiffness of the target lesion. Various types of elastography devices utilize different mechanisms. In the clinical setting, strain and shear wave elastography are mainly used for pancreatic diseases. The evaluation methods of

Role of EUS Elastography
Elastography measures the stiffness of the target lesion. Various types of elastography devices utilize different mechanisms. In the clinical setting, strain and shear wave elastography are mainly used for pancreatic diseases. The evaluation methods of strain elastography are classified into color pattern diagnosis, strain ratio (SR), and histogram analysis. Most of the reports so far have been on strain elastography, and there are still few reports on shear wave elastography [17]. The elastography pattern of PNENs is described as a homogeneous blue [18]. When comparing the strain ratio of a mass over the normal surrounding pancreatic parenchyma, malignant pancreatic masses and neuroendocrine tumors produce higher SR than inflammatory masses and normal parenchyma [19,20]. In one prospective study including six patients with a PNEN, the SR for PNENs was 56.73, higher than the SR of 17.41 seen in pancreatic adenocarcinoma [21]. In one study including malignant and benign PNENs, 67% sensitivity and 71% specificity were observed, with a malignant tumor SR cutoff of 4.4 [22]. However, reports on PNENs are still limited, and there are some problems in elastography in the diagnosis of pancreatic tumors [17]. Color pattern diagnosis is a qualitative method that depends on the operator. The cut-off values for each method and the definition of malignant/benign tumors for SR were different in each article. Shear wave elastography may overcome these limitations related to reproducibility and intra-and inter-observer variability because the hardness of the target is measured without a comparison area.

Role of Contrast-Enhanced EUS (CE-EUS)
CE-EUS is useful for the evaluation of pancreatic disease because it permits the observation of the hemodynamics of masses in real time. This technique is based on the fact that microbubbles in contrast agents are disrupted by ultrasound waves, producing signals that are detected by the ultrasound imager. Because typical PNENs have abundant blood vessels, these tumors show hypervascular contrast in the early phase, persisting until the delayed phase ( Figure 1b). CE-EUS has a high sensitivity (78.9-95.1%) and a high specificity (98.7%) in the identification of PNENs [23,24].
Ishikawa et al. reported that heterogeneous ultrasonographic texture indicates a malignant disease [23]. Furthermore, Palazzo et al. reported that contrast-enhanced harmonic EUS (CH-EUS) can accurately predict aggressive tumor behavior by evaluating the heterogeneous patterns of PNENs, with a sensitivity of 86% and a specificity of 96% [25]. Takada et al. reported that CH-EUS with time-intensity curve analysis is useful for PNEN diagnosis and grading [26].
CE-EUS plays an important role in finding a specific site within a lesion that would be more suitable for EUS-FNA. Identification of hypervascular sites in such lesions may help avoid sampling rich fibrous areas [27].

Artificial Intelligence Analysis for Endoscopic Ultrasonography
There are reports using artificial intelligence (AI) for the EUS diagnosis of pancreatic tumors. As per a recent review, there have been seven reports so far, and the diagnostic abilities in these reports were approximately 85-95% [28]. However, only one study included PNENs [29]. CE-EUS and elastography have limitations related to reproducibility and intraand inter-observer variability, but AI may be able to overcome these problems [30]. When using AI for the diagnosis of diseases, many images are required. Pancreatic tumors such as PNENs, acinar cell carcinoma, and mucinous cystic neoplasms are rarer than pancreatic ductal carcinoma and IPMN. Many EUS images of rare diseases are required to train AI to learn EUS diagnosis, but it is difficult to collect these images. However, AI will be an essential technique to help doctors to improve on their diagnostic ability using endoscopy and EUS in the near future.

Features of EUS Findings in PNENs
On EUS examination, PNENs typically appear as well-rounded, hypoechoic lesions with a homogeneous pattern and clear regular margins (Figure 1a,b). However, because PNENs grow expansively, they may cause cystic degeneration and calcification as their size increases. In these cases, PNENs often presented a heterogeneous pattern (Table 1). Table 1. Imaging findings and differential diseases of pancreatic neuroendocrine neoplasms (PNENs).

Cystic Degeneration
The most common cause of cystic degeneration is tumor bleeding, whereas tumor necrosis is rarely the cause. Cystic degeneration is mostly observed in well-differentiated PNENs. The frequency is 10-17% of all PNENs, and the larger the tumor, the higher the rate of cystic degeneration [31]. Gaujoux et al. found no association between cystic degeneration and tumor malignancy [31]. The biological behavior of cystic PNENs is somewhat less aggressive than that of solid PNENs [32,33]. Cystic degeneration is visualized as a low absorption area on contrast CT and is recognized as a nonechoic area on B-mode EUS and as an avascular area on CE-EUS. If the cysts become larger, the imaging findings will be similar to those of serous cystic neoplasms, and differentiation is necessary [34]. In EUS diagnosis, it is important to identify the solid tumor part of the cyst margin as a wall thickening or protrusion. Examples of cystic PNEN images from our department are shown in Figure 3.

Cystic Degeneration
The most common cause of cystic degeneration is tumor bleeding, whereas tumor necrosis is rarely the cause. Cystic degeneration is mostly observed in well-differentiated PNENs. The frequency is 10%-17% of all PNENs, and the larger the tumor, the higher the rate of cystic degeneration [31]. Gaujoux et al. found no association between cystic degeneration and tumor malignancy [31]. The biological behavior of cystic PNENs is somewhat less aggressive than that of solid PNENs [32,33]. Cystic degeneration is visualized as a low absorption area on contrast CT and is recognized as a nonechoic area on B-mode EUS and as an avascular area on CE-EUS. If the cysts become larger, the imaging findings will be similar to those of serous cystic neoplasms, and differentiation is necessary [34]. In EUS diagnosis, it is important to identify the solid tumor part of the cyst margin as a wall thickening or protrusion. Examples of cystic PNEN images from our department are shown in Figure 3.

Pancreatic Duct Stricture
Large PNENs may press the main pancreatic duct (MPD), resulting in stricture or obstruction; however, even small PNENs may cause MPD stenosis. It has been reported that pancreatic duct stenosis is caused not by physical compression by the tumor but by serotonin-induced stromal fibrosis [35,36]. Massironi et al. reported that CE-EUS does not show typical contrast-enhancement, that elastography shows a rigid pattern of the lesion, and that it is difficult to differentiate it from a pancreatic adenocarcinoma or an intraductal papillary mucinous neoplasm [37]. Examples of images of PNEN with MPD stenosis from our department are presented in Figure 4.

Pancreatic Duct Stricture
Large PNENs may press the main pancreatic duct (MPD), resulting in stricture or obstruction; however, even small PNENs may cause MPD stenosis. It has been reported that pancreatic duct stenosis is caused not by physical compression by the tumor but by serotonin-induced stromal fibrosis [35,36]. Massironi et al. reported that CE-EUS does not show typical contrast-enhancement, that elastography shows a rigid pattern of the lesion, and that it is difficult to differentiate it from a pancreatic adenocarcinoma or an intraductal papillary mucinous neoplasm [37]. Examples of images of PNEN with MPD stenosis from our department are presented in Figure 4.

Intraductal Invasion of the MPD
Intraductal growth of PNENs is rare. It has been reported that tumors with intraductal growth are highly malignant and have a poor prognosis [38]. Intraductal invasion shows early contrast enhancement, with a decrease during the delayed phase on CT images, and this pattern helps differentiate pancreatic PNENs from pancreatic ductal adenocarcinomas. Under EUS, if the tumor extends to the MPD, malignancy must be considered. Acinar cell carcinomas and solid pseudopapillary neoplasms (SPNs) may also show intraductal growth, and these tumors may be difficult to differentiate from PNENs.

Features of Imaging Findings in PNEN G3 and Pancreatic Neuroendocrine Carcinoma (PNEC)
Although there are histopathological differences between PNEN G3 and PNEC, their imaging findings are similar, and both of these tumors show similarities to normal pancreatic cancer (pancreatic ductal carcinoma) and pancreatic acinic cell carcinoma [39]. It has been reported that tumor blood flow correlates with prognostic factors, and the lower the vascularity, the more advanced the malignancy [40]. Histologically, the tumor does not have a capsule and grows invasively. Moreover, the tumor has abundant fibrous stroma, resulting in hypovascularity. The tumor margins are irregular, unclear, hypovascular, and there is internal necrosis of the tumor, and the above-mentioned pancreatic duct stenosis and intraductal extension occur at a high frequency. The necrotic area is recognized as a nonechoic area on B-mode EUS and as an avascular area on CE-EUS. It is difficult to distinguish using diagnostic imaging alone, and pathological examination is required. Examples of PNEN G3 images from our department are shown in Figure 5.

SPNs
SPNs are mostly seen in young female patients, and most SPNs have a good prognosis [41]. SPNs usually show characteristics similar to those of PNENs, such as solid lesions with a round shape and clear borders within the pancreas, cystic degeneration, and cystic calcification. A study comparing the EUS findings of SPNs and PNENs reported that more SPNs had a cystic component and more PNENs had hypervascularity [42]. However, differentiation is often difficult, and pathological diagnosis with immunostaining using EUS-FNA is useful for diagnosis [43,44]. Examples of images of PNENs, similar to SPNs from our department, are presented in Figure 6.

SPNs
SPNs are mostly seen in young female patients, and most SPNs have a good prognosis [41]. SPNs usually show characteristics similar to those of PNENs, such as solid lesions with a round shape and clear borders within the pancreas, cystic degeneration, and cystic calcification. A study comparing the EUS findings of SPNs and PNENs reported that more SPNs had a cystic component and more PNENs had hypervascularity [42]. However, differentiation is often difficult, and pathological diagnosis with immunostaining using EUS-FNA is useful for diagnosis [43,44]. Examples of images of PNENs, similar to SPNs from our department, are presented in Figure 6.

Serous Cystic Neoplasm (SCN)
As mentioned above, cystic, degenerated PNENs show imaging findings similar to those of macrocystic-type SCNs. In addition, solid-type SCNs show imaging findings of hypervascularity and a solid appearance and need to be differentiated from typical PNENs [34,45]. Nonenhanced CT and MRI with T2-weighted imaging and apparent diffusion coefficient maps could be helpful for the differentiation because the cystic area of PNENs shows bleeding, whereas SCNs are different in that they store serous fluid. However, these tumors are difficult to distinguish using EUS alone.

Intrapancreatic Accessory Spleen (IPAS)
IPAS is a congenital ectopic spleen that mostly occurs in the pancreatic tail. IPAS appears as a well-defined circular mass and is hypervascular with a blood flow similar to that of the spleen [46,47]. Although it is difficult to distinguish IPAS from PNEN with EUS alone, Bhutani et al. reported that careful observation shows a bridge sign connecting the lesion and the spleen [48], and Ge et al. reported that EUS-elastography is useful for distinguishing it from PNEN [49]. In T2-weighted images of superparamagnetic iron oxide MRI, IPAS has a low signal similar to that of the spleen, thus allowing it to be distinguished from PNENs. The usefulness of histological diagnosis with EUS-FNA when differentiation is difficult with diagnostic imaging has been reported [50].

Pancreatic Metastasis
Metastatic pancreatic tumors are relatively rare, and their imaging findings vary depending on the primary lesion. Renal cell carcinoma is the most common primary tumor of pancreatic metastasis. Metastatic pancreatic tumors from renal cell carcinoma become hypervascular tumors and show imaging findings similar to those of PNENs. Although information on the history of renal cell carcinoma is useful, these tumors are difficult to distinguish using imaging alone, and pathological histological diagnosis is required [51].

Serous Cystic Neoplasm (SCN)
As mentioned above, cystic, degenerated PNENs show imaging findings similar to those of macrocystic-type SCNs. In addition, solid-type SCNs show imaging findings of hypervascularity and a solid appearance and need to be differentiated from typical PNENs [34,45]. Nonenhanced CT and MRI with T2-weighted imaging and apparent diffusion coefficient maps could be helpful for the differentiation because the cystic area of PNENs shows bleeding, whereas SCNs are different in that they store serous fluid. However, these tumors are difficult to distinguish using EUS alone.

Intrapancreatic Accessory Spleen (IPAS)
IPAS is a congenital ectopic spleen that mostly occurs in the pancreatic tail. IPAS appears as a well-defined circular mass and is hypervascular with a blood flow similar to that of the spleen [46,47]. Although it is difficult to distinguish IPAS from PNEN with EUS alone, Bhutani et al. reported that careful observation shows a bridge sign connecting the lesion and the spleen [48], and Ge et al. reported that EUS-elastography is useful for distinguishing it from PNEN [49]. In T2-weighted images of superparamagnetic iron oxide MRI, IPAS has a low signal similar to that of the spleen, thus allowing it to be distinguished from PNENs. The usefulness of histological diagnosis with EUS-FNA when differentiation is difficult with diagnostic imaging has been reported [50].

Pancreatic Metastasis
Metastatic pancreatic tumors are relatively rare, and their imaging findings vary depending on the primary lesion. Renal cell carcinoma is the most common primary tumor of pancreatic metastasis. Metastatic pancreatic tumors from renal cell carcinoma become hypervascular tumors and show imaging findings similar to those of PNENs. Although information on the history of renal cell carcinoma is useful, these tumors are difficult to distinguish using imaging alone, and pathological histological diagnosis is required [51]. Examples of images of metastatic pancreatic tumors of renal cell carcinoma from our department are presented in Figure 7.

Role of EUS-FNA in PNENs
Tissue diagnosis and malignancy diagnosis using EUS-FNA are crucial for PNENs. Typical neuroendocrine tumors (NETs) need to be differentiated from SCNs, SPNs, and hypervascular pancreatic metastases. Atypical NETs, G3, and neuroendocrine carcinomas (NECs) need to be differentiated from normal pancreatic cancer and acinar cell carcinoma.

Grading Diagnosis
The grading of PNENs according to the World Health Organization (WHO) pathological classification (G1, G2, G3, and NEC according to the mitotic index and Ki-67 index) is essential for determining the treatment strategy. NEC in the WHO 2010 classification has been subdivided into G3 and NEC according to the WHO 2017 classification [64]. The treatment differed greatly between G3 and NEC.

EUS-FNA for Cystic PNENs
There are limited reports of EUS-FNA for cystic PNENs. The target of the puncture in cystic PNENs is a solid or cystic component. The cyst fluid was thin and clear, with low carcinoembryonic antigen (CEA) and amylase levels. Dhaliwal reported that the sensitivity of EUS-FNA for cystic P-NENs was 62.5%, which required FNA of both the solid and cystic components [18]. The cyst wall and septations should be targeted with FNA to maximize cytologic diagnosis [19]. In recent years, the usefulness of EUS-guided needlebased confocal laser endomicroscopy (nCLE) offering real-time microscopic imaging of the cyst epithelium providing virtual biopsies with high resolution (1-3.5 µm) has been reported [90,91]. It has also been reported to be useful in PNENs [92]. In addition, the efficacy of EUS-guided through-the-needle forceps biopsy (TTNB) for pancreatic cystic lesions has been reported [93][94][95][96]. The micro forceps, which is 0.8 mm in diameter, facilitates easy passage through a 19-gauge EUS-FNA needle and has a jaw-opening width of 4.3 mm, allowing for direct pancreatic cyst wall biopsy sampling. It is also possible to measure ki-67 in the TTNB specimens [97]. A recent study found that the combination of cyst fluid chemistry and cytology along with TTNB and/or nCLE results in a significantly higher diagnostic yield in pancreatic cystic lesions than any singular modality, although this was not statistically significant [98]. However, no large multicenter studies have specialized in cystic PNENs, even though these are necessary.

Genetic Analysis in PNENs
There are increasing reports on the use of next-generation sequencing with EUS-FNA samples in pancreatic tumors [99,100]. Recent studies have reported that alternative lengthening of telomeres, which are described as prognostic markers for resected PNENs, can be accurately performed on FNA specimens [101,102]. Genetic analysis of PNENs is also progressing [103]. A study with PNENs less than 3 cm identified genomic patterns of small PNENs associated with a different risk for liver metastases [104]. Cejas et al. elaborated on non-functional PNENs, which can predict the disease's course and can give information on postoperative clinical decisions where enhancer maps that infer gene regulatory programs were used to classify the nonfunctional PNENs [105]. Young et al. performed a comprehensive analysis of the immune response and showed that immunotherapy may be clinically beneficial for patients with the metastasislike primary (MLP)-1 subtype [106]. In addition, Simon et al. performed multi-omics on PNENs of various grades and revealed the mechanisms involved in PNENs [107]. If PNENs can be further genetically analyzed and subdivided into Ki-67 grading before surgery, it will become an attractive option for the management and preoperative risk stratification of patients with PNENs.

Conclusions
The evolution of ultrasound imaging technologies such as contrast-enhanced and elastography and the AI that analyzes them, the evolution of FNB needles, and genetic analysis will further develop the diagnosis and treatment of PNENs in the future.