Interobserver Variability and Accuracy of Preoperative CT and MRI in Pancreatic Ductal Adenocarcinoma Size Estimation: A Retrospective Cohort Study.
Can Assoc Radiol J
; 74(3): 570-581, 2023 Aug.
Article
em En
| MEDLINE
| ID: mdl-36347588
ABSTRACT
Purpose:
To assess interobserver variability and accuracy of preoperative computed tomography (CT) and magnetic resonance imaging (MRI) in pancreatic ductal adenocarcinoma (PDAC) size estimation using surgical specimens as standard of reference.Methods:
Patients with PDAC who underwent preoperative CT and MRI examinations before surgery were included. PDAC largest axial dimension was measured by 2 readers on 8 MRI sequence and 2 CT imaging phases (pancreatic parenchymal and portal venous). Measurements were compared to actual tumour size at pathologic examination. Interobserver variability was assessed using intraclass correlation coefficients (ICC) and Bland-Altman plots. Differences in tumour size (Δdiameter) between imaging and actual tumour size were searched using Wilcoxon rank sum test.Results:
Twenty-nine patients (16 men; median age, 70 years) with surgically resected PDAC were included. Interobserver reproducibility was good to excellent for all MRI sequences and the 2 CT imaging phases with ICCs between .862 (95%CI .692-.942) for fat-saturated in-phase T1-weighted sequence and .955 (95%CI .898-.980) for portal venous phase CT images. Best accuracy in PDAC size measurement was obtained with pancreatic parenchymal phase CT images with median Δdiameters of -2 mm for both readers, mean relative differences of -9% and -6% and no significant differences with dimensions at histopathological analysis (P = .051). All MRI sequences led to significant underestimation of PDAC size (median Δdiameters, -6 to -1 mm; mean relative differences, -21% to -11%).Conclusions:
Most accurate measurement of PDAC size is obtained with CT images obtained during the pancreatic parenchymal phase. MRI results in significant underestimation of PDAC size.
Texto completo:
1
Base de dados:
MEDLINE
Tipo de estudo:
Etiology_studies
/
Observational_studies
Idioma:
En
Ano de publicação:
2023
Tipo de documento:
Article