Correlation Between Apparent Diffusion Coefficient Value on MRI and Histopathologic WHO Grades of Neuroendocrine Tumors

Background: The correlation of diffusion-weighted MRI and tumor aggressiveness has been established for different tumor types, which leads to the question if it could also apply for neuroendocrine tumors (NET). Purpose: To investigate the possible correlation between apparent diffusion coefficient (ADC) value on magnetic resonance imaging (MRI) and histopathologic WHO-grades of NET. Material and Methods: Electronic patient records from patients presented at the multidisciplinary neuro-endocrine tumor board between November 2017 and April 2019 were retrospectively reviewed. Patients with both available MR imaging (primary tumor or metastasis) and known WHO tumor grade were included (n = 47). Average and minimum ADC values (avgADC; minADC) were measured by drawing a freehand ROI excluding only the outermost border of the lesion. The largest axial size (primary tumor) or most clearly delineated lesion (metastasis) was used. Results: Forty seven patients met the inclusion criteria (mean age 59 ± 12 SD; 24F/23M). Twenty one patients (45%) were diagnosed with WHO G1 tumor, 17 seventeen with G2 (36%) and nine with G3 (19%) tumor. Twenty eight primary tumors and 19 metastases were measured. A significant difference was found between low-grade (G1+G2) and high-grade (G3) tumors (Mann-Whitney; avgADC: p < 0,001; minADC: p = 0,001). There was a moderate negative correlation between WHO-grade and avgADC/minADC (Spearman; avgADC: –0,606; 95% CI [–0,773; –0,384]; minADC: –0,581; 95% CI [–0.759; –0.353]). Conclusion: Our data show a significant difference in both average and minimum ADC values on MRI between low and high grade NET. A moderate negative correlation was found between histopathologic WHO grade and ADC value.

with monoclonal anti-Ki-67 antibodies. The percentage of Ki-67 positive cells is determined in tumor hot spots where a minimum of 500 cells is counted. The mitotic index is the number of mitoses counted per high power field (HPF). Generally, mitoses are counted in 50 HPF and the mitotic index is expressed in mitoses per 10 HPF [5].
The 2017 update for pancreatic NET altered the cutoff value for NET G1 and added a subclassification of G3 tumors dividing them into well-differentiated G3 NET and poorly-differentiated G3 neuroendocrine carcinomas (NEC). The different cut-off values are demonstrated in Table 1 [4].
The diagnosis and characterization of NET is based on both laboratory testing with serum markers such as Chromogranin A (and specific hormone levels for functional NET) and multimodality imaging. Different imaging techniques are available including ultrasound (US), computed tomography (CT), magnetic resonance imaging (MRI) and functional/nuclear imaging such as somatostatin receptor imaging and positron emission tomography (PET). The combination of PET and CT (PET/CT) with different tracers can be especially valuable in NET staging and detection of metastases. Fluorine 18 fluorodeoxyglucose (FDG) PET/CT tracer is widely used in oncologic imaging but appears to be of limited value in well-differentiated NET because of the near normal glucose turnover. NET that do not show a high uptake on 18 F-FDG-PET, can be investigated with a number of somatostatin analogs labelled with Gallium 68 ( 68 Ga) (i.e. 68 Ga-DOTA-NOC) which bind to the somatostatin receptors that are expressed at the cell membrane of NET. High grade NET are more often detected by 18 F-FDG PET/CT and thus FDG avidity can be an indicator of tumor aggressiveness [6].
MRI has been used in the characterization of NET but mostly on a morphological, qualitative basis with evaluation of tumor size, borders, signal intensity, absence or presence of cystic or necrotic components, and enhancement pattern. More advanced MR imaging techniques such as diffusion-weighted imaging (DWI) and more importantly, quantitative evaluation of apparent diffusion coefficient (ADC) mapping may have an added value. The correlation of ADC values and tumor cellularity or aggressiveness/prognosis has been investigated extensively in other tumor types (i.e. prostate adenocarcinoma [7] and astrocytic brain tumors [8]) where ADC values negatively correlate with tumor cellularity and aggressiveness.
Numerous small studies indicate that similar findings may apply to NET but validation of these studies is still needed [9-18].

Purpose
To investigate the possible correlation between average and minimum ADC values of NET on MRI and the histopathological WHO grade and to determine if ADC values may help differentiate between low (G1 and G2) and high (G3) grade NET.

Case selection
For this retrospective study we included patients that were presented at the Multidisciplinary Neuroendocrine Tumor Board. This tumor board is part of a collaborative NETwork that has been set up between nine regional hospitals and the Antwerp University Hospital.
Ethical approval for this study was obtained from the Institutional Review Board (EC nr. 18/43/491) and informed consent was waived due to the retrospective nature of the study.
Elelectronic medical files from patients with a known NET discussed on the tumor board between November 2017 and April 2019 were analyzed retrospectively by a senior radiology resident and radiology staff member in consensus. Only patients with available MRI (primary tumor or metastasis), lesion size larger or equal to 1 cm, and known WHO tumor grade were included (n = 47).

Data collection and image analysis
First, the following clinical parameters were noted in an Excel worksheet: age, sex, primary tumor location, presence of metastasis, and WHO tumor grade (taken from pathology report).
The image analysis was done by a senior resident under supervision of a radiology staff member with more than ten years of experience in abdominal imaging. Measurements were performed on a picture archiving and The most notable differences of the 2010 and 2017 World Health Organization (WHO) classification system for NET is the increase of the Ki67-index cut-off value for G1 NET to <3 and the differentiation between well differentiated G3 NET and poorly differentiated G3 neuroendocrine carcinomas (NEC).
communicaion system (PACS) workstation suited for clinical use (GE RIS/PACS). Average and minimum ADC values (avgADC; minADC) were measured by drawing a freehand region of interest (ROI) on either the center slice of the lesion or the level with the least artifacts (Figures 1-3).
The T2-weighted images, DWI and contrast-enhanced T1-weighted images were used as a side by side reference.
In case of multiple lesions (i.e. liver metastases) only the largest, most clearly delineated lesion was selected. The outermost border of the lesion and cystic or necrotic regions were omitted. In very large lesions with central necrosis, the ROI was drawn in the area with the highest intensity on the corresponding high b-value DWI series. In patients with MRI of both primary tumor and metastasis, separate ROIs were drawn for each and the corresponding ADC values were noted in the Excel worksheet. Data were first sorted by WHO grade group and subsequently grouped together into low grade NET (WHO G1 and G2) and high grade NET (WHO G3) for further statistical analysis. When ADC values of both primary tumor and

Statistical analysis
Statistical analysis was conducted in SPSS Statistics (V25 -IBM). Non-parametric testing (Mann-Whitney-U) was used to compare differences in avgADC and minADC between low and high grade NET. Correlation between avgADC and minADC values and WHO grade was determined separately using R (V3.  A significant difference was found between low-grade (G1+G2) and high-grade (G3) tumors (Mann-Whitney; avgADC: p < 0.001; minADC: p < 0.001). Separate, pairwise testing of WHO grades (Kruskal-Wallis) only showed a significant difference between G1 and G3 for both avgADC and minADC (p < 0.001). There was a moderate negative correlation between WHO-grade and avgADC/minADC

Discussion
Treatment strategy for NET is related to the histological tumor grade and more specific the differentiation between low grade (WHO G1 and G2) and high grade (G3)  tumors [11]. Low grade NET can be treated with surgical resection and/or targeted therapy for example somatostatin analogs, receptor targeted radionuclide agents, bevacizumab, sunitinib, and everolimus [3,6]. High grade NET are treated with platinum-based chemotherapy. Alternative treatment strategies in inoperable patients include radiofrequency ablation, transarterial chemoembolization, and radioembolization. Both treatment strategy and prognosis are strongly dependent on histological grade with a poorer prognosis and higher metastatic rate for G3 tumors [9,11]. The possibility to predict tumor grade on a noninvasive basis and without ionizing radiation would certainly be advantageous. Early risk stratification can help with disease management and prevents treatment delay. The purpose of this study was to evaluate the correlation of ADC values on MRI with the histopathologicaly based WHO grading. Our results show a significant difference in both avgADC and minADC values between low grade (G1+G2) and high grade (G3) NETs. These findings are in line with other studies that compared G1+G2 vs G3 NET [9, 11,13,14,17,18]. We chose to compare G1+G2 vs G3 because of the different treatment and prognosis of these groups as mentioned above. However, some studies have chosen to compare G1 vs. G2+G3 NET [10,12] or compare each group separately [13,14,16,19]. When comparing differences in avgADC between separate WHO grade groups we could only find a significant difference between G1 and G3 (p < 0.001) but not between G1 and G2 (p = 0.058) or G2 and G3 (p = 0.117). This was also the case for Pereira et al. [16] (G2 vs. G3) and Besa et al.
Min et al. [15] did not compare differences between groups but found a significant (p < 0.001) moderate negative correlation between avgADC and WHO grade (-0.57), in concordance with the negative correlation of -0.61 we found and the negative correlation of -0.55 Lotfalizadeh et al. reported [14]. Besa et al.
[9] found a significant but weaker negative correlation of -0.33. Others reported a negative correlation between avgADC and Ki67-index (Guo et al. [11]: -0.41, Wang et al. [17]: -0.70). The avgADC and minADC values we found per WHO grade were similar to the values found by previous studies and are summarized in Table 3. Only the values acquired by Lotfalizadeh et al. [14] are markedly higher. The observed differences are probably multifactorial with not only