Risk stratification of gallbladder polyps larger than 10 mm using high-resolution ultrasonography and texture analysis

• Published in: European radiology 2018-01, Vol.28 (1), p.196-205
• Main Authors: Choi, Tae Won, Kim, Jung Hoon, Park, Sang Joon, Ahn, Su Joa, Joo, Ijin, Han, Joon Koo
• Format: Article
• Language: English
• Subjects: Adenoma; Adult; Aged; Aged, 80 and over; Cancer; Carcinoma; Cholecystectomy; Diagnosis, Differential; Diagnostic Radiology; Female; Gallbladder; Gallbladder - diagnostic imaging; Gallbladder Diseases - diagnostic imaging; Gallbladder Neoplasms - diagnostic imaging; Gastrointestinal; High resolution; Humans; Imaging; Internal Medicine; Interventional Radiology; Male; Medicine; Medicine & Public Health; Middle Aged; Multivariate analysis; Neuroradiology; Polyps; Polyps - diagnostic imaging; Radiology; Retrospective Studies; Risk; Sensitivity; Sensitivity and Specificity; Skewness; Texture; Ultrasonic imaging; Ultrasonography; Ultrasonography - methods; Ultrasound
• ISSN: 0938-7994; 1432-1084
• DOI: 10.1007/s00330-017-4954-1

Objectives To assess important features for risk stratification of gallbladder (GB) polyps >10 mm using high-resolution ultrasonography (HRUS) and texture analysis. Methods We included 136 patients with GB polyps (>10 mm) who underwent both HRUS and cholecystectomy (non-neoplastic, n  = 58; adenomatous, n  = 32; and carcinoma, n  = 46). Two radiologists retrospectively assessed HRUS findings and texture analysis. Multivariate analysis was performed to identify significant predictors for neoplastic polyps and carcinomas. Results Single polyp (OR, 3.680–3.856) and larger size (OR, 1.450–1.477) were independently associated with neoplastic polyps ( p  14 mm, sensitivity for differentiating neoplastic from non-neoplastic polyps was 92.3%. To differentiate carcinoma from adenoma, sessile shape (OR, 9.485–41.257), larger size (OR, 1.267–1.303), higher skewness (OR, 6.382) and lower grey-level co-occurrence matrices (GLCM) contrast (OR, 0.963) were significant predictors ( p  22 mm or sessile, sensitivity for differentiating carcinomas from adenomas was 93.5–95.7%. If a polyp demonstrated at least one HRUS finding and at least one texture feature, the specificity for diagnosing carcinoma was increased to 90.6–93.8%. Conclusion In a GB polyp >10 mm, single and diameter >14 mm were useful for predicting neoplastic polyps. In neoplastic polyps, sessile shape, diameter >22 mm, higher skewness and lower GLCM contrast were useful for predicting carcinoma. Key Points • Risk of neoplastic polyp is low in 22 mm has increased risk for GB carcinomas • Higher skewness and lower GLCM contrast are predictors of GB carcinoma • HRUS is useful for risk stratification of GB polyps >1 cm