Staging of Primary Abdominal Lymphomas: Comparison of Whole-Body MRI with Diffusion-Weighted Imaging and 18F-FDG-PET/CT

Background. The purpose of this study was to compare the accuracy of whole-body MRI with diffusion-weighted sequences (WB-DW-MRI) with that of 18F-FDG-PET/CT in the staging of patients with primary gastrointestinal lymphoma. Methods. This retrospective study involved 17 untreated patients with primary abdominal gastrointestinal lymphoma. All patients underwent 18F-FDG-PET/CT and WB-DW-MRI. Histopathology findings or at least 6 months of clinical and radiological follow-up was the gold standard. The Musshoff-modified Ann Arbor system was used for staging, and diagnostic accuracy was evaluated on a per-node basis. Results. WB-DW-MRI exhibited 100% sensitivity, 96.3% specificity, and 96.1% and 100% positive and negative predictive values (PPV and NPV), respectively. The sensitivity, specificity, and PPV and NPV of PET/CT were 95.9%, 100%, and 100% and 96.4%, respectively. There were no statistically significant differences between the two techniques (p = 0.05). The weighted kappa agreement statistics with a 95% confidence interval were 0.97 (0.95–0.99) between the two MRI readers and 0.87 (0.82–0.92) between the two methods. Conclusions. WB-DW-MRI appears to have a comparable diagnostic value to 18F-FDG-PET/CT in staging patients with gastrointestinal lymphoma.


Introduction
The gastrointestinal (GI) tract is the most common extranodal site in lymphoma, accounting for 5% to 20% of all cases [1]. Indeed, lymphomas frequently arise in the mesenteric or retroperitoneal nodes, and the abundance of lymphoid tissue in the GI tract makes this a susceptible site for secondary involvement.
Primary GI lymphomas are, however, uncommon, comprising fewer than 5% of all (GI) cancers. Indeed, according to Dawson et al. [2], a diagnosis of primary GI lymphoma should be restricted to localized disease of stages IE and IIE, whereas Lewin et al. 's [3] system requires that patients exhibit GI symptoms or a predominant lesion.
The stomach is the most common site of primary GI lymphomas, followed by the small intestine and then the ileocaecal valve [4]. The majority of such tumours (90%) are of B-cell lineage, and T-cell lymphomas and Hodgkin's lymphoma are rare. Some histological subtypes are more often found at certain locations, for example, mucosa-associated lymphoid tissue (MALT) lymphoma in the stomach, mantle cell lymphoma (MCL) in the terminal ileum, jejunum, and colon, enteropathy-associated T-cell lymphoma (EATL) in the jejunum, and follicular lymphoma (FL) in the duodenum [5]. Multifocal tumours are particularly common in MALT lymphoma and follicular lymphoma.
Accurate diagnosis and staging of primary GI lymphomas, an especially heterogeneous group of tumours, are fundamental for treatment stratification [6]. Staging of GI lymphomas is generally performed by means of Musshoff 's modified version of Ann Arbor staging [7], with the international prognostic index being used to define the prognostic subgroups. However, the system is less than optimal for documenting certain features specific to primary GI lymphoma, in particular diffuse and incurable infiltration of the GI tract. Due to this deficiency, many staging protocols and reporting systems have been proposed, and among these the Paris staging system stands out due to its ability to record the depth of tumour infiltration and specific lymphoma spread, as well as the extent of nodal involvement [8].
Various procedures are employed in diagnosis and followup and to provide data for pretreatment staging, including endoscopic ultrasound (EUS), endoscopic biopsies, computed tomography (CT), magnetic resonance imaging (MRI), 18 F-fluorodeoxyglucose positron emission tomography ( 18 F-FDG-PET), and/or molecular markers [9]. EUS and CT are the most widely performed techniques, and CT of the chest, abdomen, and pelvis exhibits high sensitivity and specificity in staging GI lymphomas. The sensitivity and specificity of CT-based staging of FL, MCL and diffuse large B-cell lymphoma (DLBCL) can be increased even further, to 80% and 90%, respectively, by the integration of 18 F-FDG-PET. However, this provides no added benefit for MALT lymphomas [10]. Moreover, EUS is considered superior to CT scan in terms of locoregional staging, as it provides details of visceral wall involvement, and for the detection of perivisceral adenopathies [11].
Whole-body diffusion-weighted MRI (WB-DW-MRI) has also been extensively studied as a method of staging lymphomas [9,12], in addition to other tumours [13,14]. Indeed, whole-body imaging techniques can be used to assess supradiaphragmatic nodal and extranodal sites, thereby providing information vital for abdominal lymphoma staging. Despite the promise being shown by WB-DW-MRI in this area, 18 F-FDG-PET/CT remains the standard of reference [14]. Nevertheless, as WB-DW-MRI does not expose the patient to ionizing radiation, its validation as a GI lymphoma staging tool would be of great clinical significance. However, to the best of our knowledge there have been no studies investigating the role of WB-DW-MRI in the staging of this type of lymphoma to date.
We therefore set out to evaluate the performance of WB-DW-MRI in staging primary abdominal GI lymphomas with respect to the current preferred method, 18 F-FDG-PET/CT. As Paris staging has not yet been universally accepted, we elected to use the modified Ann Arbor system, using histopathological findings to confirm the diagnoses.   abdominal CT scan with contrast before therapy. Diagnosis of primary GI lymphoma was confirmed histopathologically in all patients. At least 6 months of clinical and radiological follow-up was available for each patient. According to Ann Arbor staging with Musshoff 's modification, no patients were classed as stage 0 (0), two as stage I, seven as stage II, five as stage III, and three as stage IV.

Materials and Methods
All patients underwent PET/CT and MRI in the 3 weeks preceding treatment. The time interval between MRI and PET/CT scans was 0-33 days. Subsequently, 1 patient was treated by means of surgical resection and chemotherapy, while 11 received chemotherapy alone and 5 H. pylori eradication and chemotherapy. Of the treated patients, 9 achieved complete remission, 6 achieved partial remission, and 2 showed no response. Tables 1 and 2 show the clinicopathological features of patients and the histological classification and site of origin of the GI lymphomas diagnosed.
Informed consent was obtained from all patients and the local ethical committee approved this study.

Whole-Body MRI.
All patients underwent MRI on a superconductive 1.5 T magnet (Achieva, Philips, Best, Netherlands, release 2.5) using a q-body coil, with the patient positioned "feet first" on an extended anatomical coverage table featuring rolling-table technology (MobiTrak, Philips). The light visor was pointed at the orbitomeatal plane.
The PET component is a high-resolution scanner with a spatial resolution of 4.7 mm and has no septa, thus allowing 3dimensional-only acquisition. Together with the PET system, the CT scanner is used both for attenuation correction of PET data and for localization of 18 F-FDG uptake in PET images. The intravenously administrated dose of 18 F-FDG was 3,5 mBq/Kg of body weight and imaging was performed 60 minutes after administration of the tracer.
Acquired images were reconstructed using the attenuation weighted-OSEM (ordered subset expectation maximization) iterative reconstruction, with 2 iterations, 8 subsets. Fourier rebinning was used to reduce the 3D dataset to a 2-dimensional equivalent dataset, and a 4 mm full width at half maximum Gaussian filter was applied to the image after reconstruction along the axial and transaxial directions. The data were reconstructed over a 128 * 128 matrix with 5.3 mm pixel size and 2 mm slice thickness. Processed images were displayed in coronal, transverse, and sagittal planes.

Image Analysis
After acquisition and selection of the highest value, the native axial images were reformatted on a stack-by-stack basis as a single 340 mm thick maximum intensity projection (MIP) image in the coronal plane, multiple 4 mm thick multiplanar reconstructions (MPRs) in the coronal plane, and multiple 4 mm thick MPRs in the sagittal plane, oriented to include the midline as well as the spine and paraspinal regions. All reformatted images were then fused by means of the smooth fusion algorithm in MobiView software (Philips) to obtain whole-body MIP and MPR images. The grey scale was subsequently inverted to enable viewing of abnormalities (increased signal) as grey areas of varying intensity against a white background, in a PET-like visualization window, as suggested by Takahara et al. [15].
All whole-body MIP and MPR images were saved in the scanner's image database in patient-specific files. The native axial slices were separated on the basis of the value, and those acquired with = 1,000 s/mm 2 were also saved in the image database on optimal window settings. T2-STIR and SE-T1 scan data were then merged into a single whole-body image using the same method and software.
All MR and PET/CT image sets were processed by an experienced trained radiologist, anonymized, and stored in DICOM format on a CD-ROM marked with a patient-and session-specific identification code.
MR images were read independently by two experienced MRI radiologists (A.S. and A.C. with 15 and 25 years of experience in MR imaging, resp.), both blind to each other's findings and the patient's clinical status. PET/CT images were read by a specialist with 18 years of experience in nuclear medicine (G.S.).
The three readers recorded their findings on a predesigned spreadsheet (Excel, Microsoft, Redmond, USA) listing the following nodal and extranodal sites [16]: (1) Abdominal GI locations: stomach, duodenum, small bowel, colon/rectum, and multiple sites.
As Abdulqadhr et al. [17], the evaluation of possible nodal lesions was performed with MIP images of the whole-body DW sequences. The lesion was then confirmed by checking axial DW images using a value of = 1000 mm 2 /sec. A value of 0 mm 2 /sec together with = 1000 mm 2 /sec was used to rule out T2 shine-through effect [18] and to get anatomical information. The axial DW images were correlated to MIP images by references lines.
Lesions identified on DWIBS were considered positive for the disease if [18]: (1) the major axis measurement was greater than 1 cm (on axial DWIBS sequences), (2) their signal intensity on DWIBS was greater than that of the spinal cord, there were coalescent lymph nodes or nodal masses, lymph nodes were present in regions where there are normally no lymph nodes.
Central necrosis was considered a sign of malignancy, regardless of the size of the lymph node [19].
Extranodal lesions were identified as follows: (1) Presence of areas of restricted diffusion at GI tract and parenchymal organs with respect to the background signal.
(2) Correlation of the above with signal abnormalities on the morphological sequences (coronal and sagittal STIR/T1w sequences).
Bone marrow signals were considered abnormal when greater than that of muscle on T2-weighted sequences and/or when presenting a "mild" restriction of diffusion on DWIBS [20].
As apparent diffusion coefficient (ADC) values for small organs and tissues are affected by respiratory motion and fail to accurately distinguish malignant from benign lesions [21], due to partial volume effects, and since inter-and intraobserver variability afflict reproducibility of the measures [22], we neither calculated nor used ADC to characterize lesion tissues. Instead, a lesion was considered positive on 18 F-FDG-PET/CT scans in the presence of greater focal or diffuse 18 F-FDG uptake than background activity in a location incompatible with normal/physiology (or unrelated to physiological sites of tracer uptake) [23,24]. The location of 18 F-FDG uptake was always verified by CT. Lymph nodes with a shortest transverse diameter of ≥1 cm [23] or masses/coalescent lymph node were also considered positive.
The patients were staged by Ann Arbor staging with Musshoff 's modification using the data yielded by each imaging technique. The two sets of results were then compared with each other and the results of biopsy or when this was not possible with at least 6 months of clinical and radiological follow-up (CT, MRI, and PET/CT), the standard of reference. A reduction in the size of the lesion after therapy was taken as evidence that the lesion was positive for lymphoma [22].

Statistical Analysis
Sensitivity, specificity, accuracy, and positive and negative predictive values were calculated for each diagnostic method on a "per-node" basis (N-staging). Analysis of the accuracy of WB-DW-MRI, 18 F-FDG-PET/CT, and 18 F-FDG-PET without CT in the assessment of the individual disease stage of each patient was also performed. Regarding the gastrointestinal lymphoma manifestation, we only calculated the 18 F-FDG-PET/CT and WB-DW-MRI detection rates, as their large fields of view impede the accurate staging of locoregional tumour extension (stage I).
Cohen's statistics were used to calculate the interobserver agreement between the two MR readers and between WB-DW-MRI and 18 F-FDG-PET/CT. Agreement was defined as poor at < 0.2, fair at > 0.2 < 0.4, moderate at > 0.4 < 0.6, good at > 0.6 < 0.8, and very good at > 0.8.
McNemar's test was used to determine the statistical significance of differences between WB-DW-MRI and 18 F-FDG-PET/CT interpretations. A value of < 0.05 was  McNemar's test revealed no statistically significant differences between 18 F-FDG-PET/CT and WB-DW-MRI ( < 0.05).

Per-Node
The weighted kappa statistics, with a 95% confidence interval of agreement, were 0.97 (0.95-0.99) between the two MRI readers and 0.87 (0.82-0.92) between the two methods.

Per-Patient Analysis and Gastrointestinal Lymphoma
Detection Rate. Of the 17 lymphoma patients, 16 were staged the same by WB-DW-MRI and 18 F-FDG-PET/CT (94%). Of those 16 patients, 1 was classed as stage IV, and the remaining 15 were distributed between stages 0 and III (Figures 1 and 2, Table 3). In the case in which staging differed, WB-DW-MRI classed the patient as stage IV, predicting bone marrow (BM) invasion, while 18 F-FDG-PET/CT classed the patient as stage III. The bone marrow infiltration, and therefore the accuracy of the higher staging, provided by MRI, was confirmed by bone biopsy (Table 4).
In the 2 patients with low-grade MALT lymphoma, WB-DW-MRI and 18 F-FDG-PET/CT agreed (both stage II), but 18 F-FDG-PET alone predicted a far lower stage (stage 0). In the mismatched regions, enlarged lymph nodes with no F-FDG uptake were present on CT (>1 cm) (Figure 2, Table 4).    (Table 4).

Discussion
To our knowledge, this is the first study to evaluate the role of WB-DW-MRI in the N-staging of primary GI lymphoma.
Although it was not possible for us to confirm the diagnosis in all lymph nodes by histopathology (a common problem in radiological research, as it would be highly unethical to biopsy all suspected lesions) WB-DW-MRI staging was generally in full agreement with that provided by the standard of reference. In the majority of our cases, there were no differences between WB-DW-MRI and 18 F-FDG-PET/CT in this regard. However, both yielded a lower stage than the standard of reference in two patients with indolent stage 1 MALT lymphoma (Table 4), with neither PET/CT nor WB-DW-MRI being able to detect any morphologic abnormality or 18 F-FDG uptake/restriction of diffusivity in the known location of the gastrointestinal lymphoma in either of these two patients. 18 F-FDG-PET alone, on the other hand, failed to detect 18 F-FDG uptake in these and a further two cases (total 4 patients) with indolent MALT lymphoma (both stage II), which were successfully staged by integrating CT data ( Figure 2, Table 4).
As there was a high preponderance of early-stage (I-II) indolent MALT lymphoma in our group, it is perhaps unsurprising that, overall, both 18 F-FDG-PET/CT and WB-DW-MRI had a GI lymphoma detection failure rate of 40% as compared to 53% with 18 F-FDG-PET alone. Indeed, in 33 cases of MALT lymphoma, Perry et al. [25] showed 18 F-FDG-PET/CT detected active disease in 100% of advanced cases (stages III-IV) but in only 42.3% of cases of early-stage disease (I-II). It seems reasonable to assume, therefore, that the early stage, small size, and low metabolism/low 18 F-FDG uptake were behind both the downstaging of two patients and the missed gastrointestinal lymphoma manifestations. In our study, WB-DW-MRI detected one more patient with positive BM lesion than 18 F-FDG-PET/CT, correctly upstaging this patient with respect to 18 F-FDG-PET/CT. This finding is in line with those by Abdulqadhr et al. [17] showing that WB-DW-MRI correctly upstaged 3 patients with small lymphocytic lymphoma/chronic lymphatic leukaemia (SLL/CLL) as compared to 18 F-FDG-PET/CT. They postulated that the increased signal seen in SLL/CLL lesions was ascribable to the small, closely packed tumour cells causing restricted water molecule movement. However, in our two missed cases of stage 1 indolent MALT lymphoma, we failed to find any significant restriction. However, the normal focal or diffuse restriction of the small primary lesions may have been hidden by the bowel loops.
On a per-node basis, we found WB-DW-MRI to be more sensitive (100% versus 96.3%) but less specific (95.1% versus 100%) than 18 F-FDG-PET/CT. Overall, WB-DW-MRI detected a higher number of false-positive lymph nodes at the laterocervical, axillary, and inguinal nodal locations than 18 F-FDG-PET/CT, which were confirmed as reactive nodes in the follow-up. That being said, McNemar's test revealed no statistically significant differences between the two investigations, and the interobserver agreement was "very good" (>0.8).
This makes an interesting contribution to the growing debate regarding the excessive exposure to radiation during diagnostic procedures. Indeed, although 18 F-FDG-PET/CT is widely used in the management of lymphomas, it involves exposing patients to a substantial dose of radiation, which is of particular concern in young adults and children [26,27] and in patients who require repeated follow-up. A nonionizing imaging technique, such as MRI, with similar functional imaging capacity would therefore be afforded an important role in this setting, especially during follow-up.
Our results, like those of Abdulqadhr et al. [17], appear to suggest a role for WB-DW-MRI, combined with 18 F-FDG-PET/CT, in initial staging and provided that there is agreement between the two techniques as a standalone follow-up imaging technique. This would not only be safer for patients, but also improve cost-effectiveness and total examination time, both important limitations of 18 F-FDG-PET/CT, in the long term. Indeed, 18 F-FDG-PET/CT costs roughly twice as much as WB-MRI and takes considerably longer (60 minutes of waiting time after 18 F-FDG injection versus 30-35 minutes) [28].
Nevertheless, there are some limitations to our study. First and foremost, it was retrospective in nature and therefore not reliant on a standardized MRI protocol. Due to the heterogeneity of the MRI protocols adopted in our institution over the years, we only considered DWIBS and STIR/T1. However, it would have been preferable to have been able to add a WB-DW-MRI protocol with axial T1w or T2w sequences, so as to shed light on the morphological characteristics of the lymph nodes at the laterocervical, axillary, and inguinal locations, and thereby potentially eliminate false-positives. However, as no such data was available for our patients, we were unable to take them into account.
Secondly, the rarity of primary GI lymphoma meant that we were only able to consider a limited number of patients. Further studies are needed, not only to confirm our findings in a larger sample, but also to refine the staging procedure for gastric MALT lymphoma. Although indolent, this type of tumour may be multifocal [29,30], transform to DLBCL [31], and it is difficult to diagnose, endoscopic findings being normal in the majority of cases. As multiple organs are generally involved, endoscopic biopsies usually need to be taken from multiple sites of the stomach and duodenum, encompassing both normal and abnormal regions [10,32].
Our results indicate that this highly invasive procedure may one day be more accurately guided by DWIBS sequences, which could also prove useful by predicting the transformation of this tumour into higher-grade lymphoma (indeed, PET has already been recommended as a means of assessing the recurrence or transformation of various lymphomas [33]), although, once again, more research is necessary.

Conclusions
Our results suggest WB-DW-MRI as a promising technique for staging patients with primary GI lymphomas, and, pending the results of future studies and improvements to its performance, it may provide a radiation-free alternative to 18 F-FDG-PET/CT.

Disclosure
The paper has not been published elsewhere and is not currently under consideration by another journal.