Advances in the imaging of gastroenteropancreatic neuroendocrine neoplasms

Table 1 World Health Organization Classification and grading criteria for neuroendocrine neoplasms of the gastrointestinal tract and hepatopancreatobiliary organs (7)

Terminology	Differentiation	Grade	Mitoses/2 mm2	Ki-67 index
NEN grade 1	Well differentiated	Low	< 2	< 3%
NEN grade 2		Intermediate	2-20	3%-20%
NEN grade 3		High	> 20	> 20%
SCNEC	Poorly differentiated	High1	> 20	> 20%
LCNEC			> 20	> 20%
MiNEN	Well or poorly differentiated	Variable	Variable	Variable

¹Poorly differentiated neuroendocrine carcinomas are not formally graded, but are considered high-grade by definition.

LCNEC: Large cell neuroendocrine carcinoma; MiNEN: Mixed neuroendocrine non-neuroendocrine neoplasm; NEN: Neuroendocrine neoplasm; SCNEC: Small-cell neuroendocrine carcinoma.

Table 2 Sensitivity of common imaging modalities used in the evaluation of gastroenteropancreatic neuroendocrine neoplasms

Imaging modality	Sensitivity
Transabdominal USG	13%-27% for GEPNEN
Contrast enhanced ultrasound	99% in detecting liver metastases
Endoscopic ultrasonography	82%-93% for PNEN
CECT	63%-82% for PNEN
CE MRI	79% for PNEN
DWI	83% for liver metastases

USG: Ultrasonography; GEPNEN: Gastroenteropancreatic neuroendocrine neoplasms; CECT: Contrast enhanced computed tomography; CE MRI: Contrast-enhanced magnetic resonance imaging; DWI: Diffusion-weighted imaging.

Table 3 Summary of important research studies on imaging of gastroenteropancreatic neuroendocrine neoplasms

	Ref.	Number (n)	Modality	Results	Conclusion
Ultrasonography	Takada et al[25], 2019	30	Contrast-enhanced harmonic EUS	Three parameters in TIC showed high diagnostic accuracy: Echo intensity change - 87%; Rate of enhancement - 87%; Enhancement ratio for node/pancreatic parenchyma - 88.5%	Contrast-enhanced EUS and TIC analysis show high diagnostic accuracy for grading of PNEN
CT	Worhunsky et al[45], 2014	118	APCT	5-year overall survival: Hypoenhancing - 54%; Isoenhancing - 89%; Hyperenhancin - 93%. On multivariate analysis only hypoenhancement (HR 2.32, P = 0.02) was independently associated with survival	Hypoenhancement of PNEN on APCT (22% of well-differentiated PNEN) was an independent predictor of poor outcome
	Rodallec et al[46], 2006	37	Dual-phase contrast-enhanced CT	Poorly differentiated NEC: Hypoattenuating - 71%; Isoattenuating or weakly hyperattenuating - 29%; Well-differentiated NECmoderately or strongly hyperattenuating - 53%. Poor enhancement at pancreatic phase and less vascularized tumors were associated with decreased survival rate	Enhancement of PNEN at CT correlated with microscopic tumor vascularity. Low-enhancing PNEN correlated with poor differentiation and lower overall survival
	Park et al[48], 2020	69	Dynamic CT	NEC (compared to well-differentiated NEN): Significantly higher frequencies of main pancreatic ductal dilatation, bile duct dilatation, vascular invasion; Significantly lower conspicuity of interface between tumor and parenchyma, AER and PER. PER < 0.8 showed 94.1% sensitivity, 88.5% specificity for differentiation of NEC from well-differentiated NEN. On combining 3 significant CT features, the sensitivity and specificity for diagnosing NEC were 88.2% and 88.5% respectively	Tumor parenchyma enhancement ratio in portal phase is useful to distinguish NECs from well differentiated NENs. Combining qualitative and quantitative CT features aid in achieving good diagnostic accuracy in differentiation between NEC and well-differentiated NEN
	d’Assignies et al[59], 2009	36	MDCT perfusion	Tumor blood flow and intratumoral MVD showed high correlation (r = 0.620, P < 0.001). Blood flow was significantly higher in: Grade 1 than grade 2/3 tumors; Tumors with proliferation index ≤ 2% (P = 0.005); Tumors without histological signs of microscopic vascular involvement (P = 0.008). Mean transit time was longer in tumors with lymph node (P = 0.02) or liver (P = 0.05) metastasis	Perfusion CT is feasible in patients with pancreatic NENs and reflects MVD. Perfusion CT measurements correlated with histoprognostic factors, such as proliferation index and WHO grading
MRI	Canellas et al[103], 2018	80	MRI	MRI features associated with aggressive tumors: Size > 2 cm (OR = 4.8); T2 non-bright lesions (OR = 4.6); Presence of pancreatic ductal dilatation (OR = 4.9); Diffusion restriction (OR = 4.9)	MRI can assess aggressiveness of PNEN and identify patients at risk for early disease progression after surgical resection
	d’Assignies et al[74], 2013	59	MRI	DWI (71%-71.6%) was more sensitive than T2 weighted images (55.6%) and dynamic CEMRI (47.5%-48.1%). Combination of these sequences improved detection of liver metastases. Specificity of each sequence was comparable (89%-100%)	DWI is more sensitive for detection and characterization of liver metastases from NENs than T2-weighted and dynamic gadolinium-enhanced MRI
Radiomics, texture analysis and machine learning	Canellas et al[103], 2018	101	CECT with texture analysis	CT features predictive of a more aggressive tumor: Size > 2 cm (OR = 3.3); Vascular involvement (OR = 25.2); Pancreatic ductal dilatation (OR = 6); Lymphadenopathy (OR = 6.8); Entropy (OR = 3.7); Differences (P < 0.05) in progression free survival were found for: Grade 1 vs grade 2 vs grade 3 tumors; PNEN with vascular involvement; Tumors with entropy values > 4.65	CT texture analysis and CT features are predictive of aggressiveness and can be used to identify patients at risk of early disease progression after surgical resection
	De Robertis et al[77], 2018	42	MRI and histogram analysis	ADC entropy is significantly higher in grade 2/3 tumors (sensitivity: 83.3%, specificity: 61.1%). ADC kurtosis is higher in PNENs with vascular involvement, nodal and hepatic metastases (sensitivity: 85.7%, specificity: 74.3%)	Whole tumor ADC histogram analysis can predict aggressiveness in PNENs. ADC entropy and ADC kurtosis are the most accurate parameters for identification of PNEN with malignant behavior
	Luo et al[112], 2020	93	CECT with application of a CNN based DL algorithm	AUC = 0.81 of arterial phase in validation set was significantly higher than those of venous (AUC = 0.57, P = 0.03) and arterial/venous phase (AUC = 0.70, P = 0.03) in predicting the pathological grading of PNENs. The AUC and accuracy of DL algorithm for diagnosing grade 3 PNEN were 0.80% and 79.1%. There was significant difference in OS and PFS between the predicted G1/2 and G3 groups	The CNN-based DL method showed a relatively robust performance in predicting pathological grading of PNENs from CECT images
	Gao et al[114], 2019	96	CEMRI with application of deep learning algorithm on images	The average accuracy of the five trained CNNs ranged between 79.08% and 82.35%, and the range of micro- average AUC was between 0.8825 and 0.8932. The average accuracy and micro-average AUC of the averaged CNN were 81.05% and 0.8847 respectively	With the help of GAN, the CNN showed the potential to predict the grades of PNENs on CEMRI

EUS: Enhanced ultrasonography; TIC: Time-signal intensity curve; APCT: Arterial phase computed tomography; ADC: Apparent diffusion coefficient; CECT: Contrast enhanced computed tomography; CEMRI: Contrast enhanced magnetic resonance imaging; CNN: Convolutional neural network; DWI: Diffusion weighted imaging; NEC: Neuroendocrine carcinoma; NEN: Neuroendocrine neoplasm; AER: Annual equivalent ratio; PER: Portal enhancement ratio; MVD: Microvascular density; WHO: World Health Organization; MDCT: Multidetector row computed tomography; OS: Overall survival; PFS: Progression free survival; AUC: Area under the curve; PNEN: Pancreatic neuroendocrine neoplasm; GAN: Generative Adversarial Network.