Endoscopic ultrasound (EUS)-fine needle aspiration remains the gold standard for diagnosing pancreatic malignancy. However, in a subset of patients, limitations remain in regards to image quality and diagnostic yield of biopsies. Several new devices and processors have been developed that allow for enhancement of the EUS image. Initial studies of these modalities do show promise. However, cost, availability, and overall incremental benefit to EUS-fine needle aspiration have yet to be determined.
Key points
- •
Endoscopic ultrasound (EUS)-fine needle aspiration remains the gold standard for diagnosing pancreatic malignancy. However, in a subset of patients, limitations remain in regard to image quality and diagnostic yield of biopsies.
- •
Several new devices and processors have been developed that allow for enhancement of the EUS image.
- •
Initial studies of these modalities do show promise. However, cost, availability, and overall incremental benefit to EUS-fine needle aspiration have yet to be determined.
Introduction
Endoscopic ultrasound (EUS) was initially developed in the 1980s to aid in the diagnosis of abdominal pathologic abnormality. By generating transgastric and transduodenal images, EUS represented a significant advance over traditional transabdominal ultrasound or computer tomography (CT) imaging. With its ability to assess tumor size and depth, detect regional lymph node involvement and distant metastases, and also perform fine needle aspiration (EUS-FNA) to acquire tissue confirmation, EUS is now a routine part of the initial staging evaluation of a patient with suspected gastrointestinal cancer.
EUS plays an especially prominent role in pancreatic cancer. Not only does EUS provide real-time, high-resolution imaging, but it also complements imaging modalities such as helical CT and magnetic resonance imaging. EUS is particularly helpful in the evaluation of tumor resectability and the detection of vascular invasion. EUS is also exquisitely able to detect small tumors. Furthermore, the sensitivity and specificity of EUS-FNA for confirming a diagnosis of suspected pancreatic cancer have been reported as high as 92% and 96%, respectively. Pancreatic cyst fluid carcinoembryonic antigen values obtained by EUS-FNA can also be helpful in differentiating between mucinous and nonmucinous cysts.
Today, the clinical impact of EUS in patients with suspected pancreatic cancer is significant, with its use resulting in a change in diagnosis in 26% of patients and a change in management in 48% of patients undergoing an examination, including the avoidance of unnecessary surgeries. Most recently, Canto and colleagues have identified EUS as a preferred modality of initial screening for patients with an increased risk of familial pancreatic cancer and confirmed the superiority of EUS over CT for the detection of small pancreatic cysts including premalignant lesions, such as pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms, among asymptomatic high-risk individuals.
Despite the progressively increased use of EUS to detect and stage pancreatic lesions over the last 3 decades, limitations remain. Most notably, EUS depends significantly on the skill of the endosonographer. Moreover, although on-site cytopathology has been shown to improve diagnostic yield from the FNA procedure, most centers do not have the resources to provide an on-site cytopathologist or cytotechnician to confirm sample adequacy consistently. It is particularly helpful to have an onsite cytologist because obtaining a diagnosis may require as many as 5 to 7 passes in a patient with a pancreatic mass. In addition, the presence of chronic pancreatitis has been shown to limit the accuracy of EUS considerably, which is especially important, because patients with chronic pancreatitis are at increased risk for pancreatic cancer and the symptoms and imaging findings of chronic pancreatitis may closely mimic those of pancreatic cancer, engendering even more difficulty in distinguishing the 2 conditions.
As such, several general, initial improvements were developed to improve EUS as an imaging technique. The first such improvement was the development of the electronic radial echoendoscope, which studies have now confirmed provides superior image quality over the mechanical radial echoendoscope for both solid and cystic pancreatic lesions. Furthermore, unlike the initial mechanical radial echoendoscopes, the electronic echoendoscopes are compatible with the same processors as the linear echoendoscopes.
Using this electronic platform, more advanced techniques of image enhancement have been developed ( Box 1 ). This review discusses each of these advanced techniques.
Elastography
Contrast-enhanced EUS
Contrast harmonic-enhanced EUS
Three-dimensional EUS
Needle-based confocal laser-induced endomicroscopy
High-resolution microendoscopy
Optical coherence tomography
Trans-needle pancreatic cystoscopy
Introduction
Endoscopic ultrasound (EUS) was initially developed in the 1980s to aid in the diagnosis of abdominal pathologic abnormality. By generating transgastric and transduodenal images, EUS represented a significant advance over traditional transabdominal ultrasound or computer tomography (CT) imaging. With its ability to assess tumor size and depth, detect regional lymph node involvement and distant metastases, and also perform fine needle aspiration (EUS-FNA) to acquire tissue confirmation, EUS is now a routine part of the initial staging evaluation of a patient with suspected gastrointestinal cancer.
EUS plays an especially prominent role in pancreatic cancer. Not only does EUS provide real-time, high-resolution imaging, but it also complements imaging modalities such as helical CT and magnetic resonance imaging. EUS is particularly helpful in the evaluation of tumor resectability and the detection of vascular invasion. EUS is also exquisitely able to detect small tumors. Furthermore, the sensitivity and specificity of EUS-FNA for confirming a diagnosis of suspected pancreatic cancer have been reported as high as 92% and 96%, respectively. Pancreatic cyst fluid carcinoembryonic antigen values obtained by EUS-FNA can also be helpful in differentiating between mucinous and nonmucinous cysts.
Today, the clinical impact of EUS in patients with suspected pancreatic cancer is significant, with its use resulting in a change in diagnosis in 26% of patients and a change in management in 48% of patients undergoing an examination, including the avoidance of unnecessary surgeries. Most recently, Canto and colleagues have identified EUS as a preferred modality of initial screening for patients with an increased risk of familial pancreatic cancer and confirmed the superiority of EUS over CT for the detection of small pancreatic cysts including premalignant lesions, such as pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms, among asymptomatic high-risk individuals.
Despite the progressively increased use of EUS to detect and stage pancreatic lesions over the last 3 decades, limitations remain. Most notably, EUS depends significantly on the skill of the endosonographer. Moreover, although on-site cytopathology has been shown to improve diagnostic yield from the FNA procedure, most centers do not have the resources to provide an on-site cytopathologist or cytotechnician to confirm sample adequacy consistently. It is particularly helpful to have an onsite cytologist because obtaining a diagnosis may require as many as 5 to 7 passes in a patient with a pancreatic mass. In addition, the presence of chronic pancreatitis has been shown to limit the accuracy of EUS considerably, which is especially important, because patients with chronic pancreatitis are at increased risk for pancreatic cancer and the symptoms and imaging findings of chronic pancreatitis may closely mimic those of pancreatic cancer, engendering even more difficulty in distinguishing the 2 conditions.
As such, several general, initial improvements were developed to improve EUS as an imaging technique. The first such improvement was the development of the electronic radial echoendoscope, which studies have now confirmed provides superior image quality over the mechanical radial echoendoscope for both solid and cystic pancreatic lesions. Furthermore, unlike the initial mechanical radial echoendoscopes, the electronic echoendoscopes are compatible with the same processors as the linear echoendoscopes.
Using this electronic platform, more advanced techniques of image enhancement have been developed ( Box 1 ). This review discusses each of these advanced techniques.
Elastography
Contrast-enhanced EUS
Contrast harmonic-enhanced EUS
Three-dimensional EUS
Needle-based confocal laser-induced endomicroscopy
High-resolution microendoscopy
Optical coherence tomography
Trans-needle pancreatic cystoscopy
Adjunctive imaging techniques
Elastography
EUS elastography has emerged as a potentially valuable adjunctive technique in the evaluation of the solid pancreatic mass. Although traditional EUS-FNA has a high sensitivity and specificity for the detection of pancreatic cancer (92% and 96%, respectively), the yield is significantly lower in the presence of chronic pancreatitis (73.9% vs 91.3% and 54% vs 85%). This decreased yield is explained by the great difficulty endosonographers have in differentiating the inflammation of chronic pancreatitis from a discrete neoplasm. Elastography is based on the premise that malignant tissue is generally firmer than adjacent benign tissue. As such, measuring the strain generated in response to compression or vibration may be a reliable way of distinguishing benign from malignant tissues. Benign tissues would be expected to generate a larger amount of strain. Elastography has also been used extensively to evaluate the degree of liver fibrosis as well as to evaluate cervical, prostate, thyroid, and breast lesions.
Several studies have been published investigating the accuracy of EUS elastography to evaluate pancreatic masses. In their initial series of 49 patients and subsequent multicenter follow-up study of 222 patients, Giovannini and colleagues reported a sensitivity of 92.3% and a specificity of 80% of elastography to detect malignant pancreatic lesions compared with 92.3% and 68.9%, respectively, for traditional B-mode imaging. Specifically, the authors characterized elastographic images based on color and hardness, assigning each a score of 1 to 5. Although a score of 1 represented green, homogenous, soft tissue and was interpreted as normal, 5 represented blue, hard, heterogeneous tissue and was interpreted as malignancy. EUS-FNA and surgical pathologic abnormality were used to obtain the final diagnosis. Iglesias and colleagues studied 130 consecutive patients with solid pancreatic masses and 20 controls. The authors also categorized patients by specific elastographic patterns. Using a hue color map (red-green-blue), where dark blue represents hard tissue, green represents intermediate tissue, and red represents soft tissue, the 4 following groups were created: (1) green predominant, homogenous; (2) green predominant, heterogeneous; (3) blue predominant, homogenous; and (4) blue predominant, heterogeneous. Their group reported a sensitivity, specificity, and overall accuracy of EUS elastography for the diagnosis of malignancy of 100%, 85.5%, and 94%, respectively ( Fig. 1 ).
Despite these promising results, first-generation elastography has been criticized for its subjective nature. By generating a color map of tissue elastography (with blue representing the hardest tissue and red representing the softest tissue) that is superimposed on traditional gray-scale B-mode EUS images, the technique is inherently subjective; interpretation may vary between endosonographers. As such, quantitative methods, such as the strain ratio (ratio of the elasticity of a reference region in the adjacent tissue to elasticity of a given mass), hue histogram analysis, and most recently, artificial neural networks, have been developed. With regard to strain ratio, initial reports have been conflicting. Iglesias-Garcia and colleagues reported an improved specificity (92.9% vs 85.7%) and overall accuracy (97.7% vs 95.35%) of quantitative elastography over qualitative elastography for the diagnosis of malignancy among 86 patients with solid pancreatic masses; however, in the largest single-center study to date, Dawwas and colleagues showed that the area under the receiver operator curve for the detection of pancreatic malignancy using the strain ratio was only 0.69 with an associated sensitivity of 100% and a specificity of 16.7%. In terms of hue-histogram analysis, Săftoiu and colleagues have published encouraging results in a large, multicenter prospective study. By applying postprocessing software analysis to the traditional color distribution of hardness (hue histogram), the authors were able to create a quantitative scale of elasticity from 0 (softest) to 255 (hardest). Using an average hue histogram value of 175, the authors were able to show a sensitivity of 93.4%, specificity of 66.0%, and overall accuracy of 85.4% for the technique. Finally, a recent study with neural networks using artificial intelligence methodology also appears encouraging.
In an attempt to define the diagnostic accuracy of EUS elastography better, 2 recent meta-analyses were published. Mei and colleagues selected 13 studies of both qualitative (color pattern) and quantitative (strain ratio, hue histogram analysis) elastography with 1042 patients and calculated a pooled sensitivity of 95% and specificity of 67% of EUS elastography in differentiation of benign from malignant solid pancreatic masses. The area under the summary receiver operatic characteristic curve (sROC) was 0.8872, suggesting good validity of elastography as a diagnostic test. Pei and colleagues published almost identical results, with a sensitivity, specificity, and area under the sROC of 95%, 69%, and 0.8695, respectively.
In summary, EUS elastography may be a valuable adjunctive diagnostic technique in the evaluation of the solid pancreatic mass. To date, usage has been limited by its subjectivity and the need for compatible processors. Further studies will need to be performed to delineate more clearly the role of elastography in EUS and EUS-FNA.
Contrast-enhanced EUS and Contrast Harmonic-enhanced EUS
Intravenous contrast agents have long been used in transabdominal ultrasound; however, their use has only recently been applied to endoscopic ultrasound. Contrast-enhanced EUS (CE-EUS) uses either color/power Doppler EUS or dedicated contrast harmonic-enhanced EUS (CHE-EUS) as a signal intensifier. The technology is based on the finding that lesions associated with pancreatic adenocarcinoma are hypo-enhancing compared with neuroendocrine or cystic neoplasms. Although a first attempt to differentiate ductal carcinoma from chronic pancreatitis and neuroendocrine tumors was published more than 15 years ago, traditional contrast agents were too large to cross the lung bed. Newer gas-filled lipid microsphere contrast agents that are able to do so have only recently been produced.
Power Doppler EUS is a technique used to analyze the vascular abnormalities associated with pancreatic disease. By displaying the integrated power of the Doppler signal as opposed to the mean Doppler frequency shift, it is able to overcome some of the drawbacks of traditional color Doppler EUS. Even in the absence of contrast enhancement, power Doppler has been shown to be an accurate test to discriminate between pancreatic cancer and chronic pancreatitis. Săftoiu and colleagues reported an accuracy of 88% with a negative predictive value of 83% (compared with 93% and 81% for EUS-FNA) for the absence of power Doppler signals inside a mass in predicting pancreatic cancer. Furthermore, when the presence or absence of collateral vessels as detected by the power Doppler signal was factored into the analysis, the accuracy and negative predictive value improved to 95% and 92%, respectively.
The addition of intravenous ultrasound contrast agents has proved useful in conjunction with color/power Doppler in certain specific clinical contexts. First, with regards to using hypovascularity to differentiate ductal adenocarcinoma from other lesions, Dietrich and colleagues published a series of 93 patients with undetermined, solitary, and predominantly solid lesions. Using CE-EUS (color Doppler) in conjunction with the first-generation contrast agent Levovist (Bayer AG, Leverkusen, Germany), they were able to report a sensitivity and specificity for hypovascularity as a marker for malignancy in CE-EUS as 92% and 100%, respectively. All other lesions in their series (neuroendocrine tumors, serous microcystic adenomas, and 1 teratoma) revealed an either isovascular or hypervascular pattern. CE-EUS has also shown promise in differentiating chronic pancreatitis from pancreatic cancer. Hocke and colleagues published 2 studies suggesting that the addition of contrast agent SonoVue (Bracco, Amsterdam, Netherlands) to power Doppler enhances the differentiation between chronic pancreatitis and pancreatic adenocarcinoma. In their larger series of 194 patients, CE-EUS increased the sensitivity of diagnosing pancreatic cancer from 79% to 92% and the sensitivity of detecting chronic pancreatitis from 82% to 96% compared with EUS alone. Finally, in a series of 41 patients from Japan, Ishikawa and colleagues showed a higher sensitivity for CE-EUS in preoperative localization of pancreatic endocrine tumors (95.1%) compared with either multidetector CT (80.6%) and conventional transabdominal ultrasound (45.2%). On multivariable logistic regression analysis, heterogeneous ultrasonographic texture seemed to be the strongest predictor of malignancy (odds ratio 53.33, 95% confidence interval 10.79–263.58) ( Fig. 2 ).
Limitations of conventional EUS however include Doppler-related artifacts such as “overpainting” (ie, vessel and tumor size appear larger than actual secondary to larger pixel sizes) and “blooming” (ie, decreased vessel delineation secondary to overamplification of Doppler signals). As such, tissue harmonic EUS (CHE-EUS) was subsequently developed for use with ultrasound contrast agents to minimize these effects. Even without the addition of a contrast agent, tissue harmonic imaging appears to generate images that are significantly clearer than traditional EUS imaging for both cystic and solid lesions. Not surprisingly, with the addition of contrast agents, tissue harmonic imaging has shown even more promising results. Unlike power Doppler, harmonic imaging is able to display the microcirculation accurately. Napoleon and colleagues used the second-generation contrast agent SonoVue in conjunction with harmonic imaging and reported a sensitivity, specificity, and accuracy of 89%, 88%, and 88.5%, respectively, of harmonic imaging to detect pancreatic adenocarcinoma based on hypointensity compared with 72%, 100%, and 86% for EUS-FNA alone. A second group, in their series of 277 patients, reported similar results; however, they also compared CHE-EUS to multidetector row CT, and showed that CHE-EUS was significantly better than multidetector row CT in diagnosing small (less than or equal to 2 cm) carcinomas. Interestingly, Fusaroli and colleagues also reported the utility of CHE-EUS in detecting malignancy and guiding EUS-FNA in small lesions with uncertain EUS findings due to either biliary stents or chronic pancreatitis. Finally, Romagnuolo and colleagues used a second-generation perflutren lipid microsphere contrast agent to assess the accuracy of contrast-enhanced harmonic EUS. Although the positive and negative predictive values of CHE-EUS were similar to that of EUS-FNA (80.0%/100% vs 84.6%/100%), they found that in 2 of 30 cases, management changed significantly—by avoiding FNA in a liver hemangioma and a mediastinal cyst confirmed as solid.
Similar to EUS elastography, however, CE-EUS and CHE-EUS have been criticized for their qualitative nature. As such, several quantitative methods have been proposed to improve reliability. Seicean and colleagues used the index of the contrast uptake ratio to discriminate between pancreatic adenocarcinoma and chronic pancreatitis. Although both types of lesions were hypo-enhancing after standard injection with SonoVue, the index of the contrast uptake ratio was significantly lower in adenocarcinoma. Their cutoff uptake ratio index of 0.17 yielded a sensitivity of 80% and a specificity of 92%. In similar fashion, 2 groups have used the “time intensity curve” of contrast uptake and showed that the peak intensity and maximum intensity gain of lesions associated with pancreatitis were significantly higher than those associated with carcinoma.
Several prospective studies have since been published attempting to assess the diagnostic accuracy of both CE-EUS and CHE-EUS with sensitivities and specificities ranging from 92%–96% and 93%–96%, respectively, for CE-EUS to 80%–100% and 88%–99%, respectively, for CHE-EUS. In a recent meta-analysis including 1139 patients who underwent CE-EUS or CHE-HUS, the authors reported a pooled sensitivity of 94%, specificity of 89%, and area under the sROC curve of 0.9732 for assessing the accuracy of CE-EUS in diagnosing pancreatic adenocarcinoma.
In review, the addition of ultrasound contrast agents to power/color Doppler EUS is useful in certain clinical contexts. Studies have confirmed the utility of both CE-EUS and CHE-EUS in helping to differentiate pancreatic malignancy from chronic pancreatitis as well as to guide FNA in small lesions with uncertain EUS findings. As quantitative methods continue to improve and newer contrast agents continue to be developed, the use of CE-EUS may become more widespread. However, similar to elastography, CE-EUS currently remains an adjunctive technique to EUS-FNA.
Three-Dimensional EUS
Three-dimensional EUS (3D-EUS) is an emerging technique that allows for the accurate calculation of volumes by digitizing and reconstructing traditional 2-dimensional (2D) images. Each 2D image is placed into its correct location in a 3D grid, which subsequently creates a volume in cubic form. In doing so, 3D-EUS facilitates an improved understanding of spatial relationships between lesions of interest and surrounding structures. Furthermore, by allowing for the acquisition of multiplanar images, 3D-EUS may facilitate preoperative planning of particular surgical approaches. Although the data are most robust for 3D-EUS in conjunction with endorectal ultrasound in the assessment of rectal cancer and for the evaluation of anorectal fistulas, one study has shown utility with regard to pancreaticobiliary imaging. In a series of 22 patients with solid pancreatic lesions, Fritscher-Ravens and colleagues compared 3D linear EUS to conventional linear EUS to differentiate between vascular compression and vascular involvement. Using surgical histology as the gold standard, 3D-EUS proved more accurate than 2D-EUS in the evaluation of vascular involvement, particularly in patients with chronic pancreatitis. Nevertheless, given the limited amount of published data, prospective studies still must be performed to investigate the utility of 3D-EUS further in clinical pancreatic imaging.
Needle-based Confocal Laser-induced Endomicroscopy
Confocal laser-induced endomicroscopy (CLE) is an optical technique that produces images at the cellular and subcellular level, enabling the user to perform targeted, “smart biopsies,” rather than random biopsies. With adjustable imaging of plane depth, CLE yields sequential high-resolution sections through the mucosal layer. To obtain confocal images, fluorescent contrast agents are applied to a sample. An optical fiber acts as both illumination source and detection pinhole, producing high spatial resolution. The created gray-scale image is an optical section representing one focal plane within the examined specimen. Although intravenous fluoroscein is used most commonly, topical agents such as acriflavine have also been used.
The utility of CLE has been shown most widely in an array of mucosal gastrointestinal diseases such as Barrett’s esophagus, ulcerative colitis, and colorectal cancer. CLE has demonstrated the ability to differentiate neoplastic from normal tissue, and even to differentiate low-grade from high-grade dysplasia. This technology has recently become miniaturized; a new confocal miniprobe now exists (Cellvizio; Mauna Kea Technology, Paris, France), enabling easier, in vivo real-time histopathologic imaging without need for a confocal endoscope. Although probe-based CLE gives lower resolution and a fixed imaging plane depth, it has the advantage of faster image acquisition.
Recently, a needle-based miniaturized probe has broadened the potential applications to nonmucosal surfaces such as intra-abdominal viscera. The feasibility of needle-based confocal endomicroscopy (nCLE) has been evaluated in multiple animal models. Becker and colleagues first demonstrated feasibility of this needle-based system in a porcine model. Imaging was performed in 10 pigs with a confocal miniprobe inserted through a 22-gauge needle and intravenous fluorescein. Real-time images were obtained and correlated with biopsy specimens. Pancreatic imaging was deemed to be more complicated than hepatic imaging because of anatomic positioning; 8 of 10 attempts were successful. This study demonstrated that it is feasible to extend imaging into the lesion of interest and possibly enable targeted biopsy. This targeted biopsy could lead to improved yield or fewer passes during EUS-FNA; in the best scenario, it could reduce the need for onsite cytology. Similar findings were demonstrated in an EUS-guided needle-based confocal laser-induced endomicroscopy study published in abstract form. EUS-guided nCLE was again shown to be feasible. In addition, images appeared to correlate with histology.
The first study looking at the feasibility of nCLE during EUS-FNA of the pancreas in humans was authored by Konda and colleagues. The prototype was based on the CholangioFlex miniprobe (Mauna Kea Technologies). The study population included 16 cysts and 2 masses. A 19-gauge FNA needle was used to puncture the lesion. The authors report technical challenges with several features: postloading (probe loaded into needle after stylet removed), longer ferules (metallic tip at distal end of probe, protecting the device from the beveled tip of the needle), and a transduodenal FNA approach. One drawback is that this nCLE fiber required a 19-gauge needle. In another multicenter, international study by the same investigators, the authors looked at nCLE in vivo in pancreatic cysts and presented their work in abstract form. Endosonographers in 8 centers performed nCLE in 67 patients with pancreatic cysts during EUS-FNA with a 19-gauge needle. Although this study demonstrated feasibility, 2 adverse events of pancreatitis and 3 cases of bleeding occurred. Although nCLE seems promising, its current safety profile is unlikely to be acceptable to the EUS community at large.
High-resolution Microendoscopy
The authors’ group recently demonstrated the potential utility of a high-resolution microendoscope (HRME) device in pancreatic imaging. Previous ex vivo and in vivo studies in the esophagus and colon have suggested this device may be able to provide guidance to aid in performing targeted biopsy and potentially increase diagnostic yield. Although it does not have the optical sectioning ability of confocal endomicroscopy, it has demonstrated good image quality and offers the advantage of a cost-effective (less than $2500) and reusable probe. Their study demonstrated the feasibility of this new imaging technology in an in vivo swine model. Delivery of the contrast agent was straightforward and manipulation of the HRME probe was comparable to working with EUS-FNA alone. Experienced endosonographers with prior experience in HRME (over 50 cases) could achieve accuracy as high as 95% in the pancreas, indicating a relatively short learning curve. In addition, because the study coordinated the HRME images with ex vivo surgical specimens and coordinated tissue sections, an accurate pathologic diagnosis could be used as the gold standard ( Fig. 3 ). Technical issues still needing to be addressed include timing and method of dye delivery.
