New techniques in EUS: Multiparametric endoscopic ultrasound imaging

Key points

EUS has improved considerably in the past years through the development of real-time EUS elastography, contrast-enhanced EUS, and fusion EUS imaging.
Real-time EUS elastography provides qualitative and semiquantitative data about tissue stiffness, possibly allowing differentiation of benign and malignant tumors. Shear-wave elastography will certainly bring additional quantitative data.
Contrast-enhanced harmonic EUS using specific software (with low mechanical index capabilities) is already established as a procedure useful for the differential diagnosis of focal pancreatic masses.
Fusion EUS imaging represents a combination of EUS and CT/MRI, which is still under development, with the aim of decreasing the difficult learning curve of EUS but also increasing diagnostic confidence and better orientation of multiple target lesions.
Confocal laser endomicroscopy guided by EUS showed limited applicability for solid pancreatic masses, but yields reproducible results in cystic pancreatic lesions.
Artificial intelligence holds great promise, both for training but also for automated diagnosis, staging, and therapy guidance of complex pancreatic disorders.

Introduction

Endoscopic ultrasound (EUS) represents a high-resolution imaging technique used mainly for the diagnosis and staging of digestive cancers situated in the vicinity of the gastrointestinal (GI) tract. The method is increasingly used in medical centers around the world due to a significant clinical impact, especially after the addition of EUS-guided fine-needle aspiration (EUS-FNA), which is able to confirm a tissue diagnosis of malignancy. Due to the increased resolution of EUS technology, even as compared to other cross-sectional imaging methods (such as computed tomography [CT] or magnetic resonance [MR] imaging), several other methods were further developed to extend its capabilities, including real-time EUS elastography, contrast-enhanced EUS, confocal laser endomicroscopy guided by EUS and fusion imaging. ^, Last, but not least, new techniques of artificial intelligence such as deep learning are holding great promise for precise image analysis systems, which could further support automated clinical decision models based on large amounts of data.

Real-time EUS elastography

Elasticity imaging has been reported useful for the characterization and differentiation of benign and malignant tissues due to the inherent differences in the hardness of tissues. Thus, malignant tumors are usually stiffer as compared with benign masses. Consequently, the strain information induced by small tissue deformations can be computed and displayed in real time. Recently, shear-wave elastography (SWE) also completed the spectrum of ultrasound techniques. Clinical applications include breast and prostate cancers, as well as lymph nodes, thyroid masses, or focal liver lesions. Both strain and SWE were extensively used to characterize liver fibrosis in chronic liver diseases, including chronic hepatitis B or C, and also liver cirrhosis. The technique has the distinct advantage that it can be used with various ultrasound transducers, thus extending the method to virtually all organs. The method has been successfully applied with intraoperative ^, or intracavitary transducers, as well as EUS probes.

Technical details

Both the basic principles, as well as the clinical applications for the usage of ultrasound elastography, are carefully reviewed in comprehensive guidelines and recommendations issued by both the European and the World Federation Societies in Ultrasound in Medicine and Biology (EFSUMB and WFUMB). Furthermore, these guidelines were recently updated for both the liver and nonliver applications. ^,

Real-time elastography (RTE) represents a technical improvement over gray-scale ultrasound, allowing the estimation of tissue strain, during slight compressions induced by transducer or small heart/vessel movements. The method works in real time in a similar manner as color Doppler, the strain information being visually converted into a hue color scale and displayed as a transparent overlay imposed on the gray-scale ultrasound information. The principle of RTE consists of measurement of tissue displacement induced by small compressions, which are inducing strain that is usually smaller in harder tissues as compared to soft tissues ( Fig. 5.1 ). A complex algorithm called combined autocorrelation method allows the calculation of axial strain along the direction of ultrasound waves, which also corresponds to the direction of compressions. Consequently, soft tissues are easy to compress, being displayed in low-hue values approaching green, whereas hard tissues are difficult to strain, thus being displayed in high-hue values approaching blue. The information can be further quantified by taking into consideration a numerical hue scale from 0 to 255.

• Fig. 5.1, Benign mediastinal lymph node. EUS elastography showing a relatively homogeneous mixture of green and yellow, indicating a relatively soft structure as compared to the surrounding tissues (left) .

SWE is based on the usage of acoustic radiation force impulse (ARFI) technology, allowing both quantitative and qualitative data acquisition of the elasticity of a defined region of interest (ROI). SWE is based on the assessment of tissue displacement induced by acoustic shear waves of tissue vibrations generated by the mechanical impulses initiated by the ultrasound transducer. Stiffness or tissue elasticity is directly proportional with shear-wave velocity (SWV) displayed qualitatively (as color elastograms) or quantitatively (in m/s).

EUS elastography equipment includes a state-of-the-art ultrasound system, coupled with conventional endoscopic radial or linear EUS transducers. For RTE, the usual setting includes a two-panel EUS image, with the conventional gray-scale (B-mode) image on the right panel and the transparent overlay elastography image on the left side ( Fig. 5.2 ). The elastography ROI is trapezoidal in shape and can be freely selected to encompass at least half of the examined targeted lesion, as well as the surrounding tissues. Tissue elasticity values are represented in a hue color scale, with values from 0 to 255. Consequently, the color information can be semi-quantified as average values, whereas all the necessary statistical data (average strain histograms and standard deviation) can be easily calculated by using the latest versions of software ( Figs. 5.3 to 5.5 ). The system also includes the possibility of calculation of strain ratio (i.e., an estimation of the modulus ratio between two user-defined areas of interest), thus representing a semiquantitative evaluation of strain differences between the areas. However, it should be taken into consideration that changing the reference area to a deeper position significantly influences strain ratio measurements, which are otherwise independent of the size and other parameters (e.g., the elastography dynamic range). It is not yet clear if the usage of strain ratios or strain histograms should be the preferred method, while further studies will be necessary to show the differences between various methodologies. For SWE, the technology recently became available also for EUS, yielding values of SWV (m/s) and unique measurement reliability index, VsN (%). Nevertheless, an initial pilot study looking at a combination of shear-wave measurements and conventional strain elastography (based on histogram analysis) showed that SWE measurements tend to be unstable for the measurement of elasticity of solid pancreatic lesions without any significant differences between pancreatic adenocarcinoma, pancreatic neuroendocrine neoplasms, metastatic lesions, and mass-forming chronic pancreatitis (CP).

• Fig. 5.2, Malignant mediastinal lymph node. EUS elastography showing a relatively homogeneous mixture of blue, indicating a relatively hard structure as compared to the surrounding tissues (left) .

• Fig. 5.3, Malignant mediastinal lymph node. EUS elastography with EUS-guided FNA showing a relatively hard (blue) homogeneous lymph node (left) . A smaller rectangular region of interest is selected at the level of the lymph node and a hue histogram can be displayed, showing low mean elasticity (strain) values.

• Fig. 5.4, Chronic pseudotumoral pancreatitis. EUS elastography showing a relatively heterogeneous mixture of blue, green, and red, indicating a relatively intermediate elasticity structure as compared to the surrounding tissues (left) . Hue histogram analysis can be also performed to obtain semiquantitative data on the elasticity of the focal mass (mean 63.4, SD 62.3).

• Fig. 5.5, Pancreatic adenocarcinoma. EUS elastography showing a relatively homogeneous hard (blue) mass, indicating a relatively hard elasticity structure as compared to the surrounding tissues (left) . Hue histogram analysis can be also performed to obtain semiquantitative data on the elasticity of the focal mass (mean 15.8, SD 28.7).

Clinical applications

EUS RTE was initially reported to be useful in a pilot study, which included a low number of patients with focal pancreatic masses ( n = 24) and lymph nodes ( n = 25). A high sensitivity of 100% but a low specificity of 67% and 50% for pancreatic masses and lymph nodes, respectively, determined criticism of the study methodology, including qualitative pattern evaluation and establishment of diagnostic criteria in the same group of patients. The study was, however, continued with a multicenter trial that analyzed 222 patients with focal pancreatic masses ( n = 121) and lymph nodes ( n = 101), accompanied by interobserver variability data that indicated good values of the kappa coefficient of 0.785 for pancreatic masses and 0.657 for lymph nodes. EUS elastography was proven to have higher sensitivity and specificity values as compared with conventional gray-scale EUS images of 92.3% and 80.0% for the differential diagnosis of focal pancreatic masses and of 91.8% and 82.5% for the differential diagnosis of lymph nodes. Based on the published data, EUS elastography was thus suggested to be superior as compared with conventional B-mode (gray-scale) imaging that might be utilized in patients with pancreatic masses and negative EUS-FNA and also to increase the yield of EUS-FNA for patients with multiple lymph nodes. Moreover, several prospective studies using qualitative or quantitative criteria were subsequently published and supported the value of real-time EUS elastography in larger patient subgroups and multicenter trial designs ( Tables 5.1 and 5.2 ).

TABLE 5.1

Sensitivity and Specificity of EUS Elastography for the Differential Diagnoses of Lymph Nodes

Reference	Number of Lymph Nodes	Sensitivity (%)	Specificity (%)
Giovannini et al. (2006)	25	100	50
Săftoiu et al. (2006)	42	91.7	94.4
Janssen et al. (2007)	66	87.9	86.4
Săftoiu et al. (2007)	78	85.4	91.9
Giovannini et al. (2009)	101	91.8	82.5
Larsen et al. (2012)	56	55	82
Okasha et al. (2014)	88	79.3	100
Sazuka et al. (2016)	115	91.2	94.5

TABLE 5.2

Sensitivity and Specificity of EUS Elastography for the Differential Diagnoses of Focal Pancreatic Masses

Reference	Number of Patients	Sensitivity (%)	Specificity (%)
Giovannini et al. (2006)	24	100	67
Hirche et al. (2008)	70	41	53
Săftoiu et al. (2008)	43	93.8	63.6
Giovannini et al. (2009)	121	92.3	80.0
Iglesias-Garcia et al. (2009)	130	100	85.5
Iglesias-Garcia et al. (2010)	86	100	92.9
Săftoiu et al. (2010)	54	84.8	76.2
Schrader et al. (2011)	86	100	100
Săftoiu et al. (2011)	258	93.4	66.0
Dawwas et al. (2012)	111	100	16.7
Kongkam et al. (2015)	38	86.2	66.7
Kim et al. (2016)	157	95.6	96.3

Lymph nodes

An initial feasibility study that aimed to establish the value of EUS elastography for the differential diagnosis of lymph nodes was based on qualitative pattern analysis of 42 cervical, mediastinal, or abdominal lymph nodes, taking into consideration five characteristic patterns previously described for breast lesions, which allowed the establishment of a provisional diagnosis of benign ( Fig. 5.1 , ) or malignant ( Fig. 5.2 , ) lymph nodes. Sensitivity, specificity, and accuracy for the qualitative pattern analysis were 91.7%, 94.4%, and 92.86%, respectively, with an area below the receiver operating characteristic curve (AUROC) of 0.949. Several limitations of the method were acknowledged, including selection bias of the best EUS images, chosen arbitrarily by the examiner from a longer EUS elastography video. Similar results were obtained by another group that analyzed 66 mediastinal lymph nodes based on the same qualitative analysis of color patterns. The accuracy was variable for three examiners, between 81.8% and 87.9% for benign lymph nodes and between 84.6% and 86.4% for malignant lymph nodes, with an excellent interobserver analysis (kappa = 0.84).

A recent qualitative study also analyzed the role of elastography for the prediction of lymph node malignancy. Thus, consideration of different scores for the differential diagnosis indicated a sensitivity of 79.3% and specificity of 100%. For the patients with esophageal cancer only, the sensitivity and specificity of EUS elastography were 91.2% and 94.5%, respectively, significantly higher than the values of conventional B-mode EUS examinations.

Another prospective study was designed to test the accuracy of computer-enhanced dynamic analysis of EUS elastography movies for the differential diagnosis between benign and malignant lymph nodes. A total number of 78 lymph nodes were included, and average hue histograms were calculated for each EUS elastography video in order to better describe the elasticity of each lymph node according to calculations based on the hue scale of the ultrasound system. The receiver operating characteristic (ROC) analysis for the average hue histogram values inside lymph nodes yielded an AUROC of 0.928 for the differential diagnosis, with a sensitivity, specificity, and accuracy of 85.4%, 91.9%, and 88.5%, respectively, based on a cutoff level situated in the middle of the green-blue rainbow scale. The study also reported a high positive predictive value (PPV) of 92.1% and a high negative predictive value (NPV) of 85%, implying that the most probable malignant lymph nodes could be targeted by EUS-FNA ( Fig. 5.3 ), whereas EUS-FNA could be avoided in the lymph nodes that are considered most probably benign.

Another group looked at the intraobserver and interobserver agreement of EUS elastography, including the values of strain ratios, for the differential diagnosis of benign and malignant lymph nodes. Both elastography and elastography strain ratio evaluations of lymph nodes were feasible and had a good interobserver agreement of 0.58 and 0.59 (based on a cutoff of 3.81 for the strain ratio), respectively. The same group further looked at EUS elastography and elastography strain ratios based on histology results after marking of lymph nodes with EUS-FNA. The sensitivity of EUS was higher than elastography, whereas the specificity was lower as compared to elastography and strain ratios.

A recent meta-analysis that included 368 patients with 431 lymph nodes was also published, indicating a pooled sensitivity of EUS elastography of 88% with a specificity of 85% for the differential diagnosis of benign and malignant lymph nodes. After subgroup analysis with exclusion of outliers, the sensitivity and specificity were 85% and 91%, respectively, leading the authors to conclude that EUS elastography is a valuable noninvasive method used to differentiate benign and malignant lymph nodes.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here