Complementary role of computed tomography texture analysis for differentiation of pancreatic ductal adenocarcinoma from pancreatic neuroendocrine tumors in the portal‐venous enhancement phase
Christian Philipp Reinert, Karolin Baumgartner, Tobias Hepp, Michael Bitzer, Marius Horger
Abdominal Radiology (2020) 45:750–758 doi.org/10.1007/s00261-020-02406-9
Pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors (PNEN) are the two most frequently encountered pancreatic solid lesions. Approximately 60–70% of PNEN typically appear as well-defined hypervascular solid lesions in the arterial phase, which can be easily differentiated from PDAC when performing multiphasic imaging. However, PNEN can show various atypical features, including hypovascular enhancement, ill-defined margins, dilatation of pancreatic and/or biliary duct, and encasement of peripheral vessels: in this cases, the differential diagnosis with PDAC can be difficult (1).
Many patients with pancreatic lesions, however, undergo first portal-venous CT for the elucidation of the cause of cholestasis (e.g., in case of a pancreatic head mass) and of other non-specific symptoms, or the tumors are detected incidentally. Furthermore, as a progression towards malignancy is associated with derangement in vessel architecture and function, larger PNEN have a less homogenous hypervascular pattern and may show delayed contrast enhancement. Hence, pancreatic neuroendocrine tumors can be misdiagnosed as pancreatic ductal adenocarcinoma (PDAC)(2).
Therefore the aim of this study was to explore the application value of computed tomography (CT) texture analysis in differentiating pancreatic neuroendocrine tumors (PNEN) from pancreatic ductal adenocarcinomas (PDAC) in the portal-venous phase.
Texture analysis is an emerging and noninvasive method of assessing organizational characteristics, which can provide objectively quantified parameters for differential diagnosis independent of subjective analysis (3). Texture analysis can extract much more data from medical images than the naked eye by means of quantitatively analyzing greyscale distribution features, inter-pixel relations, and spatial features of images. Texture features might detect distinct quantifiable phenotypic differences of tissues which cannot be assessed through a qualitative, visual evaluation of radiological images alone. Based on these texture features, it is possible to differentiate between different tissues (4).
The study retrospectively analyzed the CT scans in the portal-venous phase (60–70 s delay) of 95 patients, 53 affected by PDAC and 42 by PNEN. Volumes of interests (VOIs) were drawn freehand by a senior radiologist with 25 years of experience in abdominal and oncologic imaging, including only the tumor tissue excluding adjacent structures. Using a commercially available software, 92 textural features were extracted including 1st, 2nd, and higher-order features, and then compared between PNEN and PDAC. Finally, radiomics analysis was used to differentiate between grades for PNEN (G2 + G3 vs G1) and PDAC (G1 vs G2 vs G3).
Compared with PNEN, PDAC had statistically lower “median”, “maximum”, “90th percentile” and “10th percentile” (p=0.0003, p=0.04, p=0.001, p=0.001, respectively). The other first-order features “energy”, “total energy” and “minimum” (p=0.00002) were significantly higher in PNEN compared to PDAC (p=0.02, p=0.0001, p=0.00002, respectively). There was no statistical significance (p >.05) for entropy between these two tumors. The 2nd order feature GLCM Imc2 was significantly higher in PDAC than in PNEN (p=0.0002). In PNEN, the higher-order feature GLSZM Small Area High Gray-Level Emphasis proved significantly higher in patients with G1 compared to patients with G2/G3 tumors (p < 05). No significant differences were found in radiomics features between different gradings of PDAC. In PNEN, the higher-order feature GLSZM Small Area High Gray-Level Emphasis was significantly higher in patients with G1 tumors compared to patients with G2/G3 tumors (p< 05). In PDAC, no significant differences were observed in textural features between patients with G1 and G2/G3 tumors. The tumor/parenchyma ratios, as well as the visual assessment into hypo-/iso-/yperdense or homogeneous/heterogeneous, did not significantly differ between PDAC and PNEN.
In conclusion, this study shows that CT-textural features quantified on portal-venous CT-image data are capable of differentiating between PDAC and PNEN, and also between G1 and G2/3 PNEN. These results may have important clinical implications as they may lead to a more individualized approach and affect clinical or surgical decisions.
REFERENCES:
[1] He M, Liu Z, Lin Y, Wan J, Li J, Xu K, et al. Differentiation of atypical non-functional pancreatic neuroendocrine tumor and pancreatic ductal adenocarcinoma using CT based radiomics. Eur J Radiol. 2019;117:102-11.
[2] Saif MW. Pancreatic neoplasm in 2011: an update. JOP. 2011;12:316-21.
[3] Incoronato M, Aiello M, Infante T, Cavaliere C, Grimaldi AM, Mirabelli P, et al. Radiogenomic Analysis of Oncological Data: A Technical Survey. Int J Mol Sci. 2017;18.
[4] Ganeshan B, Miles KA. Quantifying tumour heterogeneity with CT. Cancer Imaging. 2013;13:140-9.
Dr. Lorenzo Costa is a third-year radiology resident at University Hospital of Verona. He completed his undergraduate medical degree at the University of Bologna in 2015. He joined the Medical Imaging Department of Verona in 2017 where he is undertaking training in diagnostic and interventional radiology.
Comments may be sent to