Endoscopic ultrasound elastography in solid pancreatic lesions: individual and combined performance of strain ratio, fat-lesion strain ratio, and mean strain histogram for diagnosing pancreatic ductal adenocarcinoma

Abstract

This study aims to compare three different elastography parameters to evaluate their efficacy in diagnosing Pancreatic Ductal Adenocarcinoma (PDAC) and provide a semi-quantitative or quantitative assessment method for endoscopic ultrasound elastography. We retrospectively analyzed patients with PDAC, patients with benign pancreatic diseases, and individuals with normal pancreas who underwent endoscopic ultrasound elastography at the First Affiliated Hospital of Xi’an Jiaotong University between June 2022 and March 2023. Elastography parameters collected included strain ratio (SR), fat-lesion strain ratio (FLR), and mean strain histogram value (MEAN). Overall, 102 patients were enrolled, comprising 74 patients with PDAC (malignant group), 15 patients with benign lesions, and 13 patients with normal pancreas. Based on the median values, parameters were categorized into different groups for SR, FLR, and MEAN. A significant difference in the distribution between benign and malignant pancreatic lesions was observed across the groups (P < 0.001). ROC curve analysis for diagnosing PDAC across all study subjects showed that when SR > 15.55, FLR > 9.16, and MEAN < 23.85, the lesion area exhibited relatively lower elasticity and greater stiffness, indicating a higher likelihood of PDAC. After excluding normal pancreatic tissue, the study found that when SR > 19.145, FLR > 9.16, and MEAN < 16.25, it was easier to screen for PDAC among patients with pancreatic disease. Further combined analysis of the different parameters revealed that the combined use of the three parameters offered greater advantages in enhancing diagnostic performance. The study reveals that the strain ratio (SR), fat-lesion strain ratio (FLR), and mean strain histogram value (MEAN) are closely related to the characteristics of pancreatic diseases. SR exhibited the highest diagnostic efficacy for PDAC, and combining multiple parameters further improved diagnostic accuracy. Patients with pancreatic disease whose SR > 19.145, FLR > 9.16, and MEAN < 16.25 are more likely to be identified as PDAC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-025-01969-5.

Introduction

Pancreatic cancer presents with an insidious onset, where the majority of patients manifest no overt symptoms during the initial stages, leading to a grim prognosis with a median survival of less than one year [1, 2]. Moreover, the incidence of pancreatic cancer is increasing both domestically and internationally [3]. Early detection is crucial for the effective diagnosis and management of pancreatic cancer [4]. Once a lesion is identified, the immediate challenge is to precisely distinguish its nature [5]. For unresectable or metastatic disease, tissue confirmation is generally required because tumor molecular profiling is recommended to guide systemic and targeted therapies.

Current guidelines recommend EUS-guided fine-needle biopsy as the preferred approach to obtain tissue for histologic diagnosis of solid pancreatic lesions. The procedure has a low complication rate, and the risk of needle-tract seeding is very low [6–8].However, establishing an accurate diagnosis of solid pancreatic lesions remains challenging in specific settings, including chronic pancreatitis and cases in which prior EUS-guided sampling is nondiagnostic or negative [9–11].

Notably, meta-analyses report high pooled sensitivity but only modest specificity for diagnosing pancreatic malignancy with EUS elastography, therefore, elastography should not be used as a stand-alone diagnostic test in routine practice.Against this backdrop, Endoscopic ultrasound (EUS) elastography serves as an adjunct to standard EUS-guided tissue acquisition [12–14].It applies gentle external compression to induce tissue strain, which the system renders as a color map overlaid on the ultrasound image. It provides real-time, relative stiffness information and helps direct biopsy toward firmer regions within a lesion for targeted sampling [15, 16]. Evidence also indicates that stiffer pancreatic masses may require more needle passes to achieve a comparable diagnostic yield, supporting pass-number adjustment according to stiffness [17]. Because qualitative color mapping is operator dependent and subjective, semi-quantitative and quantitative measurements were developed and are now used in practice [18–22].In chronic pancreatitis, where inflammatory change and heterogeneous echotexture make differentiation difficult, elastography offers semi-quantitative or quantitative stiffness measurements that help define suspicious areas and select the biopsy site [23, 24].

Although many studies have examined the role of EUS elastography in differentiating pancreatic nodules, acquisition and interpretation are not fully standardized, and no universally accepted cut-off values exist across devices and metrics.Building on these gaps, we retrospectively evaluated the strain ratio (SR), the fat-lesion strain ratio (FLR), and the mean strain-histogram value (MEAN), comparing their individual and combined performance for diagnosing pancreatic ductal adenocarcinoma.

Materials and methods

Study design

The retrospective single center study was approved by the Ethics Committee of the First Affiliated Hospital of Xi’an Jiaotong University (Approval No. XJTU1AF2023LSK-525).Informed consent was obtained from all participants prior to enrollment. All procedures were performed in accordance with the ethical standards of the Declaration of Helsinki. Clinical management followed Pancreatic Adenocarcinoma, Version 2.2021, NCCN Clinical Practice Guidelines in Oncology [25].

Patients

All patients who underwent endoscopic ultrasound elastography (EUS-EG) at the First Affiliated Hospital of Xi’an Jiaotong University between June 2022 and March 2023 were identified through the institutional electronic medical record and endoscopy reporting system.Patients were first categorized by index EUS findings into two groups:

In the pancreatic disease group, a solid focal pancreatic lesion was visualized on EUS.EUS-guided tissue acquisition (FNB) was performed during the EUS examination to obtain pathological diagnosis in line with current guideline recommendations for solid pancreatic lesions [8]. Malignant lesions were defined as ductal adenocarcinoma of the pancreas (PDAC), while benign lesions were defined by the absence of malignant cells and the presence of inflammatory cells on pathology. For analysis, cases were classified into malignant lesions and benign lesions; the benign category included entities such as chronic pancreatitis.

The normal pancreas group consisted of individuals who underwent EUS evaluation due to clinical indications (such as nonspecific upper abdominal symptoms or suspicious imaging findings), but in whom EUS revealed no pancreatic abnormalities. No tissue sampling was performed in the normal pancreas group because no focal pancreatic lesion was identified.

For the pancreatic disease group, inclusion criteria were: (1) Patients underwent ultrasound-guided FNB during endoscopic ultrasound examination; (2) Endoscopic ultrasound confirmed that the solid pancreatic lesion was located within the pancreatic parenchyma; (3) Availability of histopathology from FNB (or surgical specimen when available), allowing final classification as malignant or benign; (4) No known history of another primary malignancy outside the pancreas.Exclusion criteria were: (1) Lesions arising outside the pancreatic parenchyma; (2) Pathology indicating neuroendocrine tumor, intraductal papillary mucinous neoplasm, or other malignancy; (3) Missing elastography data for strain ratio (SR), fat-lesion strain ratio (FLR), or mean strain-histogram value (MEAN).

EUS elastography procedure

All EUS procedures were performed using a Hitachi ultrasound imaging system with a compatible echoendoscope (Hitachi, Japan). An experienced endosonographer (with over 500 prior EUS cases) performed all procedures and obtained all elastography measurements. A routine ultrasound scan of the pancreas was then performed through the gastric wall. Upon locating the lesion, strain elastography mode was activated. The echoendoscope tip was kept in gentle, steady contact with the gastric or duodenal wall, and light external pressure was applied to generate tissue deformation. The system displayed a real-time, color-coded elasticity map superimposed on the grayscale image, reflecting relative tissue stiffness (harder tissue showing less strain).When a stable elastographic pattern had been maintained for approximately 3–5 s without major motion artifact, a still frame was captured.A region of interest (ROI) was then manually drawn to include the entire solid component of the lesion while avoiding obvious vessels, ductal structures, or cystic/necrotic areas.

Elastography parameters and measurement

We recorded three elastography parameters at the time of the index endoscopic ultrasound: strain ratio, fat lesion strain ratio, and mean strain histogram value. The system stored these values in the institutional reporting database, and a data manager exported the fields to a de-identified dataset for analysis. Elastography metrics were abstracted from the reporting system under outcome-masked conditions. Below we define each metric, the regions of interest (ROIs), and their interpretive significance (Fig. 1a–c).

Fig. 1.

Measurement of each parameter: (a) Strain Ratio (SR): In a representative elastography image, the pancreatic lesion is outlined as region A (red circle) and an adjacent area of normal pancreatic tissue is outlined as region B (blue circle). (b) Fat-to-Lesion Strain Ratio (FLR): In the same image, the lesion is designated as region L (red circle) and an area of peripancreatic fat is designated as region F (blue circle). (c) Mean Strain Value (MEAN): A rectangular region of interest (ROI, red box) is manually placed by the examiner over the lesion or tissue of interest. The ultrasound system then computes the average (mean) strain of all pixels within this ROI. The mean strain value thus provides a single numerical measure of the tissue’s elastic deformation within the selected ROI

Strain Ratio (SR): The strain ratio (SR) is defined as the quotient of the strain in region B divided by the strain in region A. This semiquantitative metric compares tissue deformation in normal parenchyma versus the lesion.

Fat-Lesion Strain Ratio (FLR): The fat-to-lesion strain ratio (FLR) is calculated as the strain in region F divided by the strain in region L. This ratio quantifies the relative stiffness of fat tissue compared to the lesion.

Mean Strain Histogram (MEAN): This value is automatically calculated by the ultrasound software based on the region of interest (ROI) selected by the examining physician.

Statistical methods

Data analysis was conducted using IBM SPSS Statistics version 27.0. For normally distributed continuous variables, the mean ± standard deviation (Mean ± SD) was utilized; for skewed continuous variables and ordinal data, the median was employed. The chi-square test was used to compare categorical variables. The diagnostic capability of elastography values in distinguishing PDAC was evaluated by plotting receiver operating characteristic (ROC) curves, determining the area under the curve (AUC), sensitivity, specificity, and the optimal cutoff value, thus assessing their clinical relevance. A p-value of less than 0.05 was established as the threshold for statistical significance.

Results

Patients’ characteristics

Between June 2022 and March 2023, a total of 223 patients underwent endoscopic ultrasound elastography at the First Affiliated Hospital of Xi’an Jiaotong University. Following the application of inclusion and exclusion criteria, 102 patients were eligible for final analysis, including 89 individuals with pancreatic lesions and 13 with a normal pancreas (Fig. 2). Among those with pancreatic lesions, pathological diagnosis identified 74 patients with PDAC (malignant group) and 15 patients with benign lesions, comprising autoimmune pancreatitis (n = 7), chronic pancreatitis (n = 7), and granulomatous inflammation (n = 1). The baseline clinical characteristics of all participants—including those in the malignant, benign, and normal pancreas groups—are detailed in Table 1.

Fig. 2.

Patient Inclusion Process

Table 1.

Clinical characteristics of lesions and normal pancreas

	Malignant Lesions (N = 74)	Benign Lesions (N = 15)	Normal Pancreas (N = 13)
Characteristics
Age Range	31–88	30–83	36–78
Mean Age ± Standard Deviation	63.69 ± 10.49	59.53 ± 11.34	63.08 ± 10.87
< 60 years	26 (35.1%)	7 (46.7%)	3 (23.1%)
≥ 60 years	48 (64.9%)	8 (53.3%)	10 (76.9%)
Gender
Female	33 (44.6%)	2 (13.3%)	5 (38.5%)
Male	41 (55.4%)	13 (86.7%)	8 (61.5%)
CA19-9 > 39U/mL	65 (87.8%)	7 (46.7%)	7 (53.8%)
CA125 > 35U/mL	36 (48.6%)	4 (26.7%)	10 (76.9%)
CEA > 5U/mL	32 (43.2%)	1 (6.7%)	2 (15.4%)
Symptoms
Abdominal Pain	60 (81.1%)	7 (46.7%)	0 (0.0%)
Gastrointestinal Symptoms	31 (41.9%)	8 (53.3%)	0 (0.0%)
Jaundice	23 (31.1%)	9 (60.0%)	0 (0.0%)
Asymptomatic Coexisting Diseases	6 (8.1%)	2 (13.3%)	0 (0.0%)
Past medical history
Hypertension	18 (24.3%)	5 (33.3%)	0 (0.0%)
Coronary Heart Disease	3 (4.1%)	2 (13.3%)	0 (0.0%)
Diabetes	17 (22.9%)	5 (33.3%)	1 (7.7%)
Social History
Smoking History	18 (24.3%)	11 (73.4%)	1 (7.7%)
Drinking History	9 (12.2%)	5 (33.3%)	1 (7.7%)

Factors of various elastography parameters

The normality tests indicated that the distributions of SR, FLR, and MEAN values were non-normally distributed. Therefore, patients were grouped according to the median of each parameter (SR: 27, FLR: 18, MEAN: 13). Given that our data were binary categorical, we employed the chi-square test to analyze the differences among the groups. The chi-square results (Supplementary Tables 1–3) revealed significant differences in the distribution of benign and malignant pancreatic lesions across different SR, FLR, and MEAN groups (P < 0.001). Additionally, CEA level > 5 U/ml was significantly associated with high SR values (P = 0.022), high FLR values (P = 0.002), and low MEAN values (P < 0.001). Furthermore, a correlation was observed between CA19-9 levels and FLR values (P = 0.033) as well as MEAN values (P = 0.002). A history of alcohol consumption was associated with SR values (P = 0.029) and FLR values (P = 0.006), while a history of diabetes was significantly correlated with low MEAN values (P = 0.033).

Diagnostic efficacy of SR, FLR, and MEAN for PDAC in the overall study population

Univariate chi-square analysis revealed significant associations between PDAC and the distributions of SR, FLR, and MEAN values among all study participants. ROC curves illustrating the diagnostic performance of SR, FLR, MEAN, and their combinations for distinguishing PDAC from benign lesions and normal pancreatic tissue are shown in Fig. 3a - d. Table 2 summarizes the diagnostic parameters, including area under the ROC curve (AUC), sensitivity, specificity, and optimal cutoff values. Among the individual parameters, SR demonstrated the highest AUC (0.870), indicating superior overall diagnostic accuracy. The FLR value showed slightly lower sensitivity but higher specificity compared to SR, suggesting fewer false positives. Conversely, MEAN exhibited the highest sensitivity but the lowest specificity, indicating fewer false negatives but more false positives. Optimal cutoff points determined from ROC analysis were as follows: SR > 15.55, FLR > 9.16, and MEAN < 23.85, which were predictive of PDAC. Pairwise combinations of these parameters showed improved diagnostic efficacy compared with single parameters. Specifically, combinations involving SR (SR + FLR, SR + MEAN) exhibited higher sensitivity than FLR + MEAN, while the combination FLR + MEAN demonstrated higher specificity. Notably, combining all three parameters (SR + FLR + MEAN) provided the greatest AUC, reflecting optimal diagnostic accuracy.

Fig. 3.

ROC curves illustrating the diagnostic performance of elastography parameters (SR, FLR, MEAN) and their combinations in distinguishing PDAC from benign pancreatic lesions and normal pancreatic tissues in the overall study population. (a) ROC curve using SR values. (b) ROC curve using FLR values. (c) ROC curve using MEAN values. (d) ROC curves using pairwise combinations (SR + FLR, SR + MEAN, FLR + MEAN) and combined parameters (SR + FLR + MEAN)

Table 2.

Diagnostic performance of ROC curve analysis for PDAC in the combined cohort (Including PDAC, benign pancreatic Diseases, and normal pancreatic Tissue)

	AUC	Sensitivity	Specificity	Optimal Cut-off Value
Elastic Imaging Parameters
SR	0.87	0.905	0.75	15.55
FLR	0.837	0.851	0.786	9.16
MEAN	0.829	0.946	0.643	23.85
SR + FLR	0.875	0.919	0.75	/
SR + MEAN	0.873	0.919	0.75	/
FLR + MEAN	0.839	0.851	0.78	/
SR + FLR + MEAN	0.874	0.919	0.75	/

Diagnostic efficacy of SR, FLR, and MEAN for PDAC in the patients with pancreatic disorders

Receiver Operating Characteristic (ROC) curves were constructed to evaluate the diagnostic performance of SR, FLR, MEAN, and their combinations in differentiating PDAC from benign pancreatic diseases among patients with pancreatic disorders (Fig. 4a-d; Table 3).Among individual elastography parameters, SR exhibited the highest area under the ROC curve (AUC = 0.800) and highest sensitivity (85.1%), indicating superior diagnostic efficacy. FLR showed comparable sensitivity (85.1%) and specificity (66.7%) to SR but had a slightly lower AUC (0.771). Conversely, MEAN had the lowest diagnostic accuracy with an AUC of 0.710, a sensitivity of 79.7%, and specificity of 66.7%. Pairwise combinations of parameters demonstrated distinct trade-offs between sensitivity and specificity. Although the combination of SR + FLR achieved the highest AUC (0.816) and highest specificity (80.0%), it showed the lowest sensitivity (68.9%), implying an increased risk of missed diagnoses. The combination SR + MEAN had the highest sensitivity (91.9%) but lowest specificity (60.0%), indicating potential for increased false-positive diagnoses. The FLR + MEAN combination showed intermediate values across all diagnostic indicators (AUC = 0.774; sensitivity = 81.1%; specificity = 66.7%). Notably, the simultaneous combination of all three parameters (SR + FLR + MEAN) maximized the diagnostic performance, achieving the highest AUC (0.826) along with improved specificity (73.3%) and satisfactory sensitivity (81.1%).

Fig. 4.

ROC curves demonstrating the diagnostic efficacy of elastography parameters (SR, FLR, MEAN) and their combinations in differentiating PDAC from benign pancreatic diseases among patients with pancreatic disorders. (a) ROC curve using SR values. (b) ROC curve using FLR values. (c) ROC curve using MEAN values. (d) ROC curves using pairwise combinations (SR + FLR, SR + MEAN, FLR + MEAN) and combined parameters (SR + FLR + MEAN)

Table 3.

Diagnostic performance of ROC curve analysis for PDAC in the pancreatic diseases (Including PDAC, and benign pancreatic Diseases)

	AUC	Sensitivity	Specificity	Optimal Cut-off Value
Elastic Imaging Parameters
SR	0.8	0.851	0.667	19.145
FLR	0.771	0.851	0.667	9.16
MEAN	0.71	0.797	0.667	16.25
SR + FLR	0.816	0.689	0.8	/
SR + MEAN	0.81	0.919	0.6	/
FLR + MEAN	0.774	0.811	0.667	/
SR + FLR + MEAN	0.826	0.811	0.733	/

Optimal cutoff values and clinical implications

Based on ROC curve analysis performed on the entire study population, optimal diagnostic cut-off values were determined: SR > 15.55, FLR > 9.16, and MEAN < 23.85 indicate that the lesion has lower elasticity, greater stiffness, and is more likely to be PDAC. Conversely, SR < 15.55, FLR < 9.16, and MEAN > 23.85 indicate relatively higher elasticity, reduced stiffness, and a greater likelihood of benign conditions.After excluding data from patients with normal pancreatic tissue, optimal cut-off values specific to differentiating PDAC from benign diseases among patients with pancreatic disorders were adjusted as follows: SR > 19.145, FLR > 9.16, and MEAN < 16.25 indicate a higher likelihood of PDAC, whereas SR < 19.145, FLR < 9.16, and MEAN > 16.25 suggest benign pancreatic diseases.

Discussion

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal malignancies of the gastrointestinal tract [26], largely because it is often detected at an advanced stage and remains difficult to distinguish from benign inflammatory masses at the time of presentation [27, 28]. Endoscopic ultrasound–guided fine-needle biopsy (EUS-FNB) is considered the reference method for obtaining a histologic diagnosis in solid pancreatic lesions because it offers high diagnostic accuracy, low complication rates, and core tissue suitable for molecular profiling in unresectable disease [29–31]. However, its diagnostic yield can fall in technically difficult scenarios such as a background of chronic pancreatitis, very small (≤ 10 mm) lesions in the pancreatic head, or cases in which a prior EUS-guided pass was nondiagnostic; in these settings, false-negative results and repeated sampling remain a concern. These challenges create a clear role for adjunctive, real-time techniques like EUS elastography, which can highlight the stiffest and most suspicious subregion of a lesion to guide needle targeting rather than replace tissue acquisition.

Previous clinical investigations have demonstrated that in inflammatory lesions, the activation of pancreatic stellate cells results in the secretion of significant quantities of collagen, leading to varying degrees of fibrosis and subsequent tissue hardening in the affected region [32, 33]. In the progression of malignant tumors, the disruption of normal differentiation control mechanisms leads to uncontrolled proliferation of tumor cells, which lack proper organization and tissue architecture, resulting in the formation of tightly packed cell clusters that enhance tissue stiffness [34, 35]. Simultaneously, tumor cells alter the surrounding extracellular matrix to facilitate their growth and dissemination, secreting copious amounts of extracellular matrix components like collagen and fibronectin, thereby contributing to increased tissue rigidity. By quantifying tissue elasticity through various parameters, it may be feasible to distinguish the differences in stiffness between malignant tumors and benign masses.

The application of ultrasound elastography in the diagnosis and management of pancreatic diseases commenced with real-time tissue elastography (RTE). The first generation of RTE relied solely on color representation for the qualitative assessment of tissue stiffness in the examined area. Uchida et al. conducted RTE examinations on 53 cases of pancreatic tumors, demonstrating that the diagnostic accuracy of conventional B-mode ultrasound alone ranged from 70% to 80%, while the integration of B-mode ultrasound with RTE surpassed 90%, thereby underscoring the diagnostic potential of elastography in pancreatic disease [36]. However, owing to the subjective nature of this method, quantitative diagnostics utilizing the strain ratio were introduced with the advent of second-generation RTE. Studies have shown that quantitative techniques, including the strain ratio and fat-lesion strain ratio, significantly improve diagnostic accuracy when applied to breast masses. The application of quantitative methods in the diagnosis of pancreatic space-occupying lesions can enhance the differential diagnostic utility of elastography for PDAC.

The Strain Ratio (SR) is presently the most widely utilized semi-quantitative technique for diagnosing the nature of pancreatic diseases; however, various studies have reported differing assessments of its diagnostic efficacy. Iglesias-Garcia reported that the SR for pancreatic malignant tumors was significantly higher than for inflammatory masses, with a sensitivity of 100% and specificity of 92.9% for diagnosing pancreatic malignancies, yielding an area under the curve (AUC) of 0.983 [37]. In contrast, Dawwas et al. observed a sensitivity of 95.7% for detecting pancreatic malignant tumors using SR, but with a specificity of only 16.7% and an AUC of 0.69 [38]. These discrepancies might be attributed to variations in clinical experience and operational practices among different ultrasound physicians, which influence the selection of reference area B in SR calculations. Consequently, SR results from different operators are not directly comparable and exhibit limited reproducibility. In this study, SR values were calculated by experienced senior ultrasound physicians who consistently selected non-lesion pancreatic tissue at the same horizontal level as the lesion’s center, thus ensuring high credibility.

Concerning the optimal cutoff value for SR, researchers have reported a range of values. Okasha found that setting the SR cutoff value at 7.8 for EUS-E resulted in a sensitivity of 98%, specificity of 77%, accuracy of 92%, PPV of 91%, and NPV of 95% for distinguishing between benign and malignant pancreatic lesions [39]. Altonbary demonstrated that with an SR cutoff value of 7.75, EUS-E achieved a specificity of 99.9%, sensitivity of 90.7%, accuracy of 92.2%, PPV of 99.9%, and NPV of 67.9% in diagnosing benign versus malignant pancreatic lesions [40].

Unlike previous studies that have predominantly focused on a single parameter (such as SR), our study systematically include and compare the diagnostic performance of three parameters: the strain ratio (SR), the fat-lesion strain ratio (FLR), and the strain elastography histogram mean (MEAN). This analysis clarified the relative advantages of each parameter in identifying PDAC, with SR demonstrating the most prominent overall diagnostic performance. Additionally, by comparing diagnostic indicators such as AUC, sensitivity, and specificity across different populations, we found that when the control group included individuals with normal pancreases, the diagnostic performance of all parameters was significantly better than in scenarios where only benign lesions were used as controls. This suggests that EUS elastography has a stronger discriminative ability between lesions and normal tissue, whereas it faces greater challenges in distinguishing benign from malignant lesions. This disparity has important methodological implications for the design of future studies and clinical screening strategies. On this basis, we further evaluated the diagnostic value of different combinations of parameters. The results showed that the combined use of all three parameters provided greater advantages in enhancing sensitivity and diagnostic stability. This finding suggests that, in clinical interpretation, the complementarity among parameters should be emphasized, and combined analysis may improve diagnostic accuracy. Notably, FLR, as a semi-quantitative parameter based on adipose tissue, inherently has a more fixed reference region and lower operator dependence, which results in better consistency and reproducibility in multi-operator or multi-center settings. This indicates that FLR has the potential to be integrated into standardized screening protocols. Finally, we also found that the performance of the MEAN parameter varies across different diagnostic tasks. When distinguishing pancreatic lesions from normal pancreas, MEAN exhibited the highest sensitivity, but its diagnostic efficacy in differentiating benign from malignant lesions was relatively weak. Based on this, our study is the first to propose that MEAN is more suitable for initial lesion screening rather than qualitative diagnosis, providing a more refined reference for clinical parameter selection.

Our study has certain limitations. The sample size is relatively small and uneven, as it is a single-center retrospective study. We included an inadequate number of cases, particularly in the benign lesion subgroup (including patients with chronic pancreatitis) and the normal pancreas control group, which may limit the statistical power in assessing the diagnostic performance of the three elastography parameters. This small sample size, especially for benign conditions, could compromise our ability to fully evaluate SR, FLR, and MEAN in differentiating benign from malignant lesions. We have acknowledged this as a limitation and caution that our findings should be interpreted accordingly. We also recommend that future studies with larger cohorts — including more patients with benign tumors and chronic pancreatitis — be conducted to validate and extend our results.

In our post-hoc exploratory analysis restricted to the PDAC subgroup, elastography parameters did not differ by location (head/neck vs. body/tail) or by a 40-mm size dichotomy (all P > 0.05) (Supplementary Tables 4,5). This analysis does not assess very small lesions nor the incremental diagnostic yield of elastography-guided sampling. In particular, small pancreatic head lesions (≤ 10 mm) pose a known diagnostic challenge for EUS-FNA, as obtaining adequate tissue from such tiny lesions can be difficult and may lead to higher false-negative rates. In these challenging cases, EUS elastography could provide additional benefit by highlighting the stiffest (hardest) areas of a lesion to target for needle sampling. Preliminary studies suggest that elastography-guided targeting may improve sampling performance for small or technically challenging lesions, but this requires confirmation in larger cohorts. We have added this discussion point to the revised manuscript to acknowledge the potential value of lesion size and site considerations. Future research should specifically evaluate different lesion sizes and locations (especially tiny pancreatic head lesions) to determine whether elastography-guided FNA can further improve diagnostic accuracy in such cases. This would help to address the questions raised about whether, for a pancreatic head lesion < 10 mm, EUS-elastography could augment the accuracy of EUS-FNA.

Nevertheless, considering that endoscopic ultrasound is an invasive procedure, the practicality of multiple examinations by different physicians is constrained. Thus, when selecting reference areas, multiple measurements can be made, and an average value can be calculated to yield more stable SR data. Pancreatic malignancies extend beyond PDAC; pancreatic neuroendocrine tumors (NETs) and intraductal papillary mucinous neoplasms (IPMNs) may also exhibit malignancy. Nonetheless, owing to significant histopathological differences among NETs, IPMNs, and PDAC, their tissue elasticity might also vary considerably. As a result, this study was focused exclusively on PDAC, assessing the diagnostic performance of SR, FLR, and MEAN for this particular malignancy.

In summary, this study provides novel evidence and insights across multiple dimensions—including parameter comparison, combined optimization, control group selection, and task-specific applicability. Our findings not only expand the evidence base for EUS elastography in the diagnosis of pancreatic cancer but also offer practical guidance for its future standardization and clinical translation.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Fangzhou Wang and Jian Ma contributed equally to this work. Fangzhou Wang, Jian Ma, Zheng Wu were responsible for the study concept and design. Fangzhou Wang conducted the primary analysis, collected data and drafted the manuscript. Jian Ma, Haonan Liu, Cancan Zhou, Jie Hao and Shuai Wu contributed to the formulation of the experimental protocol and acquisition of data. Jiaoxing Wu, Haonan Liu, Ruiqi Cao and Zhengyuan Feng collaboratively engaged in the analysis of data and the writing and revision of the manuscript. Hao Sun, Zheng Wang, Weikun Qian reviewed submitted version of manuscript and supervised the study. Zheng Wu critically assessed the study, revised and finalized the manuscript.

Data availability

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.