Differentiation between pancreatic metastases from clear cell renal cell carcinoma and pancreatic neuroendocrine tumor using double-echo chemical shift imaging.
Abdom Radiol (NY). 2018 Oct;43(10):2712-2720. doi: 10.1007/s00261-018-1539-7.
Lyu HL1, Cao JX2, Wang HY3, Wang ZB4, Hu MG5, Ma L6, Wang YW6, Ye HY6.
PURPOSE: The purpose of the study was to retrospectively analyze whether double-echo gradient-echo (GRE) chemical shift imaging (CSI) can differentiate between pancreatic metastases from clear cell renal cell carcinoma (PM-ccRCC) and pancreatic neuroendocrine tumor (pNET).
METHODS: Institutional review board approval and informed consent were waived. CSI, T2WI, DWI, and DCE magnetic resonance imaging (MRI) were performed in patients with PM-ccRCC and pNET. Eleven patients with PM-ccRCC and 24 patients with pNET were enrolled into this retrospective study. The signal intensity was measured in the pancreatic tumor and spleen on in-phase and opposed-phase images. The signal intensity index (SII) and tumor-to-spleen ratio (TSR) in PM-ccRCC and pNET were calculated and compared. Receiver operating characteristic (ROC) analysis was performed to evaluate the diagnostic accuracy of SII and TSR in the differentiation between PM-ccRCC and pNET.
RESULTS: The SII between PM-ccRCC and pNET (20.3% ± 16.8% vs. - 3.2% ± 11.4%) was significantly different (P < 0.001), as was the TSR (- 19.2% ± 16.6% vs. 6.0% ± 13.8%) (P < 0.001). The area under the ROC curve was 0.917 for the SII and 0.902 for the TSR. Additionally, an SII threshold value of 8.1% permitted the differentiation of PM-ccRCC from pNET with a sensitivity of 90.9%, a specificity of 91.7%, a positive predictive value of 90.1%, a negative predictive value of 91.7%, and an accuracy of 91.4%. A TSR cut-off value of - 4.7% enabled the differentiation of the two groups with a sensitivity of 79.2%, a specificity of 90.9%, a positive predictive value of 90.9%, a negative predictive value of 79.2% and an accuracy of 82.9%.
CONCLUSION: Double-echo GRE chemical shift MR imaging can accurately differentiate between PM-ccRCC and pNET.