Accuracy of apparent diffusion coefficient in differentiating pancreatic neuroendocrine tumour from intrapancreatic accessory spleen

Objectives: To evaluate and compare the accuracy of absolute apparent diffusion coefficient (ADC) and normalised ADC (lesion-to-spleen ADC ratio) in differentiating pancreatic neuroendocrine tumour (NET) from intrapancreatic accessory spleen (IPAS). Methods: Study included 62 patients with the diagnosis of pancreatic NET (n=51) or IPAS (n=11). Two independent reviewers measured ADC on all lesions and spleen. Receiver operating characteristics (ROC) analysis to differentiate NET from IPAS was performed and compared for absolute and normalised ADC. Inter-reader reliability for the two methods was assessed. Results: Pancreatic NET had significantly higher absolute ADC (1.431x10^-3 vs 0.967x10^-3 mm²/s; POpenSPiltSPi0.0001) and normalised ADC (1.59 vs 1.09; POpenSPiltSPi0.0001) compared to IPAS. An ADC value of ≥1.206x10^-3 mm²/s was 70.6% sensitive and 90.9% specific for the diagnosis of NET vs. IPAS. Lesion to spleen ADC ratio of ≥1.25 was 80.4% sensitive, and 81.8% specific while ratio of ≥1.29 was 74.5% sensitive and 100% specific in the differentiation. The area under the curve (AUCs) for two methods were similar (88.2% vs. 88.8%; P=0.899). Both methods demonstrated excellent inter-reader reliability with ICCs for absolute ADC and ADC ratio being 0.957 and 0.927, respectively. Conclusion: Both absolute and normalised ADC allow clinically relevant differentiation of pancreatic NET and IPAS. Key points: • Imaging overlaps between IPASs and pancreatic-NETs lead to unnecessary procedures including pancreatectomy. • Uniquely low ADC of spleen allows differentiating IPASs from pancreatic NETs. • Both absolute-ADC and normalised-ADC (lesion-to-spleen ADC-ratio) demonstrate high accuracy in differentiating IPASs from NETs. • Both methods demonstrate excellent inter-reader reliability.