Authors
Ming-Sheng Chien, Ting-Chung Wo, Jian-Han Lai
Published in
BMC gastroenterology. Jul 15, 2026. Epub Jul 15, 2026.
Abstract
Endoscopic ultrasound (EUS) is a highly sensitive imaging modality for detecting pancreatic tumors, particularly small lesions. However, its ability to distinguish malignant from benign lesions remains a diagnostic challenge. This study aimed to evaluate the imaging characteristics observed via EUS in differentiating malignancy and to assess the influence of tumor size on diagnostic accuracy.
This retrospective study included 315 patients who underwent EUS evaluation for pancreatic tumors between January 2019 and July 2024. The final diagnosis was established based on surgical pathology, EUS-FNB results, or clinical follow-up ≥ 6 months. Key imaging features assessed included echogenicity, lesion margins, pancreatic duct dilatation, and distal pancreatic atrophy. Tumors were categorized by size (< 2 cm or > 2 cm), and statistical analyses were conducted to identify predictive factors of malignancy.
Of the 315 tumors, 205 (65.1%) were malignant. Five EUS imaging features-hypoechoic pattern, heterogeneous echotexture, irregular margins, pancreatic duct dilatation, and distal pancreatic atrophy-were identified as predictive of malignancy. The ROC-derived cutoff of 3.5 corresponds clinically to ≥ 4 features. AUC was 0.807 (95% CI: 0.759-0.849). Using a ≥ 4-feature threshold, sensitivity was 87.8%, specificity 54.5%, PPV 78.6%, and NPV 69.4%. Subgroup multivariable analysis revealed that heterogeneous echotexture and pancreatic duct dilatation were significant predictors of malignancy in tumors < 2 cm. However, in tumors > 2 cm, no individual feature significantly distinguished malignancy.
Among pancreatic tumors, those exhibiting four or more of the five EUS features should be prioritized for malignancy evaluation. In small tumors, heterogeneous echotexture and pancreatic duct dilatation offer valuable diagnostic insight. Conversely, larger tumors present greater diagnostic challenges due to overlapping features on B-mode imaging alone.
PMID:
42458314
Bibliographic data and abstract were imported from PubMed on 16 Jul 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 12
- Comments 0