Validation of in vivo nodal staging of solid malignancies with USPIO-enhanced MRI: a workow protocol

In various cancer types, the rst step towards extended metastatic disease is the presence of lymph node metastases. Imaging methods with sucient diagnostic accuracy are required to personalize treatment. Lymph node metastases can be detected with ultrasmall superparamagnetic iron oxide (USPIO)-enhanced magnetic resonance imaging (MRI), but this method needs validation. Here, a workow is presented which is designed to compare MRI-visible lymph nodes on a node-to-node basis with histopathology.


Abstract Background
In various cancer types, the rst step towards extended metastatic disease is the presence of lymph node metastases.
Imaging methods with su cient diagnostic accuracy are required to personalize treatment. Lymph node metastases can be detected with ultrasmall superparamagnetic iron oxide (USPIO)-enhanced magnetic resonance imaging (MRI), but this method needs validation. Here, a work ow is presented which is designed to compare MRI-visible lymph nodes on a node-to-node basis with histopathology.

Methods
In patients with prostate, rectal, periampullary, esophageal, and head-and-neck cancer, in vivo USPIO-enhanced MRI was performed to detect lymph nodes suspicious for harboring metastases. After lymphadenectomy, but before histopathological assessment, a 7 Tesla (T) preclinical ex vivo MRI of the surgical specimen was performed, and in vivo MR-images were radiologically matched to ex vivo MR images. Lymph nodes were annotated on the ex vivo MRI for an MR-guided pathological examination of the specimens.

Results
Matching lymph nodes of ex vivo MRI to pathology was feasible in all cancer types. The annotated ex vivo MR images enabled a comparison between USPIO-enhanced in vivo MRI and histopathology which allowed for analyses on nodal, or at least on nodal station level.

Conclusions
A work ow was developed to validate in vivo USPIO-enhanced MRI with histopathology. Guiding the pathologist towards lymph nodes in the resection specimens during histopathological work-up allowed analysis at nodal level, or at least nodal station level of in vivo suspicious lymph nodes with corresponding histopathology, providing direct information for validation of in vivo USPIO-enhanced MRI detected lymph nodes.
Trial registration VALINODE data is collected from patients undergoing USPIO-enhanced MRI in the clinical setting. Therefore, the trial has not been registered in an online trial register. 7TNANO1 is registered at Clinicaltrials.gov: NCT02751606, April 26 2016

Background
Many different types of solid cancers have a high propensity to metastasize to locoregional and distant lymph nodes which is one of the most important prognostic factors for survival [1,2]. Current clinical practice employs considerable overtreatment for precautionary reasons. If an accurate pre-treatment lymph node status is established, therapeutic strategies can be tailored to the individual patient. With continuously improving opportunities to selective treatment of individual metastatic deposits with surgery or radiotherapy, assessment and exact localization of lymph node metastases is crucial. Concurrently, increased certainty of the absence of lymph node metastases could safely minimize treatment. Therefore, highly sensitive detection of small lymph node metastases is urgently needed.
Compared to other imaging techniques, magnetic resonance imaging (MRI) provides superior soft-tissue contrast. This allows good visualization of small anatomical structures and can be used to characterize lymph nodes [3]. MRI in combination with a contrast agent of ultrasmall superparamagnetic particles of iron oxide (USPIO) might be a promising, generalized, diagnostic tool to detect lymph node metastases in various types of cancers [4]. In short, 24-36 hours after intravenous administration, normal lymph nodes accumulate USPIO nanoparticles in macrophages causing a strong attenuation of the MR signal, while suspicious lymph nodes without USPIO retain MR signal on T2* weighted sequences [5]. Although promising results were previously reported in different body regions [6], the clinical value of USPIO-enhanced MRI using state-of-the-art 3-dimensional (3D) sequences with high isotropic spatial resolution should be appropriately validated [7,8]. This demands for an accurate validation of USPIO-enhanced MRI compared to histopathology on a nodal level.
Few attempts have been undertaken to match individual lymph nodes on MRI with the corresponding lymph node on histopathology after lymphadenectomy [9,10]. Although results were promising, correlation of lymph nodes on a nodeto-node level remains di cult and small lymph nodes (< 3mm) often cannot be correlated with pathology. A seven Tesla (T) ex vivo MRI might enable to visualize the detailed anatomy of the resected specimen with the locations of dissected lymph nodes. In this study, we developed a protocol to correlate in vivo visible lymph nodes with histopathology data on a node-to-node level using 7T ex vivo MRI of lymphadenectomy specimens as an intermediate step. This is a way to validate in vivo MRI with histopathology on a nodal level to improve nodal staging.

Subjects
The data is collected from one trial investigating ex vivo MR scanning of the resection specimen in prostate cancer patients who underwent USPIO-enhanced MRI in the clinical setting and from four trials that investigate the feasibility of USPIO-enhanced MRI for detecting suspicious lymph nodes (rectal cancer NCT02751606, pancreatic and periampullary cancer NCT04311047, esophageal cancer NTR6072, and head-and-neck cancer NCT03817307). To illustrate the feasibility of the proposed work ow, one patient for each of the ve types of solid malignancies is presented.
Characteristics of all patients are described in table 1. All studies were approved by the central ethics committee on research involving human subjects and the local ethics committee Arnhem-Nijmegen. Signed informed consent was obtained before patient inclusion according to each corresponding research protocol. In all patients, 2.6 mg/kg body weight of USPIO (Ferumoxtran-10, Ferrotran® SPL Medical B.V. Nijmegen, the Netherlands) mixed with 0.9% saline solution was administered intravenously with a slow-drip infusion of 30-45 minutes under continuous clinical supervision. 24-36 hours later an MRI examination was performed using a 3T MR system (Magnetom Skyra or Prisma, Siemens Healthcare, Erlangen, Germany). Sequences were focused on detecting USPIO in lymph nodes and consisted of a 3D T1-weighted volumetric interpolated breath-hold examination (VIBE) Dixon pulse sequence and 3D multi gradient echo (MGRE) sequence [11]. The MRI pulse sequence parameters are described in tables 2 and 3. All patients, except for the head-and-neck cancer patient, received intramuscular smooth muscle and bowel relaxant (Scopolamine Butyl Bromide, Sano -aventis Netherlands B.V.) to reduce peristaltic motion. The periampullary cancer patients also received glucagon to reduce peristaltic motion,  Flip angle (°) 10 9 9 10 10 Abbreviations: 3D = 3-dimensional, FOV = eld of view, Hz = Hertz, min = minutes, ms = milliseconds, TE = echo time, TR = repetition time, VIBE = volumetric interpolated breath-hold examination. Flip angle (°) 10 10 10 10 10 Abbreviations: 3D = 3-dimensional, FOV = eld of view, Hz = Hertz, MEDIC = multiple echo data image combination, mGRE = multi-gradient echo sequence, min = minutes, ms = milliseconds, n.a. = not applicable, TE = echo time, TR = repetition time.

Surgery and ex vivo MR examination
Resection of the primary tumor and lymphadenectomy was performed according to routine clinical procedure.
Immediately after resection, the specimen was transported at room temperature from the operating room to the Department of Pathology. The surgical specimen was, according to each different study protocol, either xated with formalin before (prostate, rectal, esophageal, and head-and-neck cancer)) or after ex vivo MRI (periampullary cancer). This ex vivo MRI was executed on a 7T preclinical MR system using a volume radiofrequency coil (ClinScan, Bruker® BioSpin, Ettlingen, Germany). The specimen was placed in a proximal-distal position in the preclinical MR system ( gure 8). The MR protocol consisted of a T1-weighted 3-dimensional GRE sequence with frequency-selective lipid excitation and a 3D multi-GRE pulse sequence with ve acquired echoes and frequency selective water excitation. The ex vivo MR sequence parameters are described in table 4. All surgical specimens with a size exceeding 10 cm were scanned in multiple contiguous sections and were eventually merged to one composed MR image dataset. MeVisLab software (MeVis Medical Solutions, Fraunhofer MEVIS) was used to view and annotate the 3D ex vivo composed lipid and waterselective images simultaneously. All structures that were spherical and bordered in three dimensions were identi ed as lymph nodes. Abbreviations: 3D = 3-dimensional, min = minutes, MRI = magnetic resonance imaging, ms = milliseconds, TE = echo time, TR = repetition time.
The interval between USPIO administration and surgery varied between 1-9 days (table 1). In our experience, USPIO remains visible on MR-images until 2-3 days after infusion. The effect of iron particle accumulation in healthy (parts of) lymph nodes is only visible when the ex vivo MRI is performed within this timeframe.
MRI guided pathology and node-to-node correlation In vivo MR-images were evaluated by one or two radiologist(s) and the researcher coordinating the study. Based on lymph node size and appearance, and their location in relation to anatomical landmarks, these in vivo nodes were correlated to the ex vivo MRI images. Subsequently, the lymphadenectomy specimen was dissected at the Department of Pathology with the 3D ex vivo MR images presented on-screen to guide the pathologist [12]. The matched lymph nodes were located in the specimen, taken out, and separately enclosed in a tissue cassette. The lymph nodes were numbered accordingly on in vivo MRI, ex vivo MRI and the corresponding tissue cassette. All lymph nodes were, when larger than 5.0 mm, sectioned multiple times in 2.0-3.0 mm thick sections for para n embedding. All nodes were examined with a hematoxylin and eosin (H&E) stain on 4 μm sections. Relating the histopathological slide to the corresponding ex vivo assigned lymph node and subsequently to the corresponding in vivo identi ed lymph node provided the complete node-to-node correlation.

Results
In vivo USPIO-enhanced MRI and ex vivo MRI of resected specimens were feasible in all ve patients. The general work ow to improve nodal staging for solid malignancies of different anatomical origins is schematically visualized in gure 1. Due to different clinical routines in pathological work-up for each primary tumor localization, each tumorspeci c work ow currently developed has differences in approach. These speci cs steps are outlined below per primary tumor localization.

Prostate cancer
A bilateral extended pelvic lymph node dissection (PLND) was performed in the patient with prostate cancer for diagnostic purposes before treatment with radiotherapy. Each anatomical nodal station is separately dissected and pinned onto the corresponding region of a schematic anatomical plate ( gure 3). In order to preserve anatomical orientation, each nodal station is colored differently ( gure 3) and left-and right specimens are separately vacuum sealed using the TissueSAFE plus device (Milestone, Sorisole, Italy, gure 9). Vacuum sealing the specimens offers several advantages: the separate specimens of each nodal station are scanned in their original con rmation, scanning the separate specimens together ensures that enough MR signal is generated, and their orientation is preserved hence routine clinical practice is not disturbed. An ex vivo MRI of the vacuumed specimens is performed. The specimens are subsequently processed routinely at the dissection room; ex vivo MR-images are not present during this process. The results of both the in-and ex vivo MRI are correlated to the nal histopathological results on a level-to-level basis. In case ex vivo MRI yields a higher number of lymph nodes than histopathological analysis, additional dissection of the specimen is performed to determine if the number of harvested evaluated lymph nodes can be increased.

Rectal cancer
In the rectal cancer patient, surgical resection was performed according to the principle of a total mesorectal excision (TME) [13]. Lymph nodes that are likely to be involved in rectal cancer are located in the mesorectum, which is the fat tissue surrounding the bordered on the outside by the mesorectal fascia. This fascia forms an important anatomical barrier for tumor spread and is the intended surgical boundary of the TME procedure. Matching the ex vivo specimen to the in vivo appearance of the mesorectum after a well-performed surgical procedure is therefore relatively easy. Pathological assessment of the resected specimen is performed according to the method described by Quirke et al [14].
The specimen is evaluated from distal to proximal by slicing serial transverse tissue lamina with a 5 mm interval. All nodes detected on pathology were correlated to the ex vivo MR-images which could be correlated to the in vivo MRimages for a node-to-node analysis ( gure 4).

Periampullary cancer
In the patient with periampullary cancer a pancreaticoduodenectomy with lymphadenectomy was performed. Most lymph nodes were en-bloc resected with the tumor, additionally several lymph nodes were resected separately. After resection, the surgeon pinned the specimen onto an anatomical drawing displaying the pancreatic region including its lymph node stations of which a digital photograph was taken ( gure 5). At the department of Pathology, the resection planes are inked, the primary tumor is incised and the entire specimen is xated with formalin [15]. The xated specimen is processed routinely according to the axial slicing technique ( gure 5) [15]. Nodal correlation is performed on a level-tolevel basis and, where possible, on a node-to-node basis.

Esophageal cancer
For the esophageal cancer patient, esophagectomy with two eld lymphadenectomy (thoracic and abdominal lymph node stations) was performed after neoadjuvant chemoradiotherapy. During surgery, the esophagus including locoregional lymph nodes were en-bloc dissected. The surgeon labelled all the lymph node stations within the surgical specimen with colored plastic beads, according to the TIGER study protocol [16]. Differences in con guration of the ex vivo specimen as compared to its original anatomical location makes it di cult to correlate the lymph node stations using anatomical landmarks. However, using the labelled location of the lymph node stations, lymph nodes on in vivo MRI are matched to ex vivo MRI and histopathology on a nodal, or at least a nodal station, level ( gure 6).

Head-and-neck cancer
In the head-and-neck cancer patient, the neck dissection specimen was taken out en-bloc. The primary tumor was resected separately. After resection, the specimen was pinned to a grid with a reference drawing of the neck including its anatomical nodal levels and xated with formalin ( gure 7). Next, the different neck levels are separated. In order to maintain orientation, these pieces are vacuum sealed in their original con guration with a TissueSAFE plus system ( gure 9) and scanned ex vivo in the 7T preclinical MR system. The cuts, or the boundaries between the neck levels, are visible on ex vivo MRI. All suspicious nodes were individually matched on in vivo MRI, ex vivo MRI and histopathology. These nodes are cut into multiple sections and stained with immunohistochemistry in addition to standard hematoxylin and eosin staining. All non-suspicious nodes are routinely dissected and processed per level. Therefore, this work ow enabled node-to-node correlation for suspicious nodes and level-to-level correlation for non-suspicious nodes.

Discussion
We developed a method which enabled correlation of in vivo detected lymph nodes from different primary solid malignancies with pathology. The work ow was technically feasible in prostate, rectal, periampullary, and head-andneck cancer patients. The intermediate step of the 7T ex vivo MRI scan of resected specimens allowed in principle correlation on a nodal level, but at least on a nodal station level between the in vivo MRI images and histology. For each tumor location, different approaches were required to maintain anatomical orientation of the resected specimen and correctly match the results to the pathological evaluation. In esophageal, periampullary, and rectal cancer, lymph nodes were dissected en-bloc with the primary tumor. Therefore, lymph nodes could be matched using corresponding anatomical landmarks to the ex vivo MRI. In head-and-neck and prostate cancer, lymphadenectomy and primary tumor dissection were performed separately. Hence the lymphadenectomy specimens were vacuum sealed in plastic bags in order to maintain orientation and easy positioning of the specimen within the scanner. Korteweg et al examined healthy axillary lymphoid tissue of two deceased females on a clinical 7T ex vivo MRI system and detected lymph nodes <1 mm, corroborating our results [3]. Both studies [10,3] proposed a framework for exact matching of radio-pathological ndings by pinning the dissected specimen to an MR compatible grid containing an explanatory absciss and ordinate [10] or vertical and horizontal reference lines [3]. For each lymph node, the in-plane position was assessed and correlated to the MR images. Matching was successful in 80% [10] and 88% [3] of the lymph nodes, although this percentage strongly depends on the total amount of identi ed lymph nodes.

Strengths and limitations
We developed a work ow to match individual lymph nodes detected on in vivo MRI to the resection specimen by incorporating 7T ex vivo MRI to the protocol. This method can be used in studies validating MRI, such as USPIOenhanced MRI, for detection of lymph node metastases on a nodal level with histopathology as gold standard. The method was illustrated in prostate, rectal, periampullary, esophageal, and head-and-neck cancer. The results demonstrate that various approaches enabled maintenance of orientation in both large, small and anatomically complex resection specimens. Therefore, the presented work ow can easily be adapted to other organs. High resolution ex vivo MRI images provide direct information on the location of lymph nodes in resection specimens and is of immediate help to harvest nodes in general, or suspicious nodes in particular. In case of a short time interval between USPIO-enhanced MRI and surgery, maintenance of signal intensity on the ex vivo MR-images represents the absence of USPIO deposits and thus a lymph node suspicious for harboring a metastasis. This functional information can be bene cial for node-to-node correlation. Radiologic differentiation between a lymph node and a blood vessel was formerly potentially di cult. However, using 3D high-resolution MR sequences enabling visualization of nodal structures in the transverse, sagittal and coronal plane, this is no longer an issue.
Some potential limitations should also be mentioned. Node-to-node correlation for all lymph nodes is challenging.
Differences in spatial resolution between in and ex vivo exams can lead to differences in detection of lymph nodes, particularly the small ones. Additionally, there could be a difference between the resected volume and a volume de ned as resected on in vivo images, which also can lead to differences in the amount of evaluated lymph nodes. Also, the xation process needed for the pathology processing causes volume change of the specimen, which leads to a change of con guration of the specimen and landmarks. Neck dissection and PLND specimens, for example, are relatively small but harbor many lymph nodes and generally contain only a few anatomical landmarks. Therefore, it is of paramount importance that the anatomical orientation of the specimen is well documented during dissection. In this way a reliable node-to-node correlation can be achieved or in case of clusters of small lymph nodes a per nodal station correlation can be performed. In addition, the ex vivo MRI requires a considerable investment of time which is not always desirable in clinical practice. Likewise, not every institution has an 7T preclinical MR system available for ex vivo measurements. Therefore, the method we developed is particularly suitable for studies validating imaging techniques for nodal assessment.

Clinical implications
In various types of cancer, detection of small lymph node metastases is urgently needed. Both over-and undertreatment in oncological therapeutic regimens result in unnecessary morbidity and mortality. Treatment needs to be tailored on a patient level to overcome this problem. Thus, knowledge regarding merely the presence (N+) or absence (N0) of nodal metastases is not su cient anymore. Clinicians need to be informed about the number, size and exact location of metastatic deposits which is a prerequisite to target therapies. Individual treatment strategies consist of local tumor excision or ablation without lymphadenectomy in N0 patients, reducing unnecessary morbidity in this group of patients.
In N+ cancer patients the exact N-staging can lead to additional treatment options for local lymph node metastases such as surgical resection, ablation, or image guided radiotherapy. Currently, a variety of promising non-invasive diagnostic imaging tests are developed to detect lymph node metastases, e.g. hybrid PET-MRI, USPIO-enhanced MRI and targeted uorescence imaging [17,4]. Using our described work ow with 7T ex vivo MRI as an intermediate step to guide histopathological work-up, new techniques can be validated on a node-to-node level with histopathology. The detailed 7T ex vivo MRI of the resection specimen aids in nding the smallest lymph nodes that otherwise might be missed with routine pathological evaluation [12], supporting the validation of in vivo imaging techniques.

Conclusion
A technically feasible method was developed to correlate lymph nodes identi ed on USPIO-enhanced MRI for the detection of lymph nodes metastases with histopathology of the same nodes as gold standard. This work ow was suitable in various anatomical regions. Ex vivo MRI-guided pathological dissection enabled lymph node correlation on a nodal or station level between in vivo MR-images and histopathology. This approach allows a validation of in vivo USPIO-enhanced MRI for evaluation of loco-regional lymph nodes which is essential for a subsequent diagnostic accuracy assessment. Consent to participate -all patients included provided informed consent for study participation.

Abbreviations
Consent for publication -all patients included provided informed consent for their data to be published anonymously.
Availability of data and material -The data that support the ndings of this study are available upon reasonable request from the corresponding author DD. The data are not publicly available due to them containing information that could compromise research participant privacy. Figure 1 Schematic overview of the work ow Example of tissue xation in head-and-neck cancer. A grid containing a reference drawing of the neck including the levels (a) is present at the operating room. After resection, the fresh specimen is pinned to this grid (b) and is subsequently xated with formalin at the department of Pathology (c) Figure 3 Example of the work ow in prostate cancer. In vivo MRI with a suspicious lymph node in the right obturator fossa (a, white dashed lined box). Each nodal station was colored and pinned to an anatomical reference drawing for preservation of orientation (b). The surgical specimen was vacuum sealed to xate the specimen for the ex vivo MRI (c).

Figures
Due to a short USPIO-MRI interval of 1 day, USPIO particles were still present in the resection specimen. Since there was no loss of signal intensity in this lymph node, it was regarded suspicious (d, white dashed lined box). Pathology assessment with hematoxylin and eosin staining showed that this lymph node was metastatic (e) Figure 4 Example of the work ow in rectal cancer. A coronal view of the in vivo MRI showing the rectum within the white dashed lined box with a non-suspicious lymph node located right pararectally close to the rectal wall (a). After TME surgery the specimen was xated in formalin (b) of which an ex vivo MRI scan was performed. After the ex vivo scan, the specimen is sliced in 5 mm thick lamina slices (c) which can be compared to the transverse images of the ex vivo MRI (d). The lymph node was detected on ex vivo MRI as well as on pathology. Since the USPIO-surgery interval was 9 days, and USPIO was no longer present in the resection specimen, there were no changes in signal intensity. For this reason, the lymph node has maintained signal on the water-selective ex vivo MR-image (d). Histopathological assessment with hematoxylin and eosin staining revealed a non-metastatic lymph node (e) Figure 5 Example of the work ow in periampullary cancer. A transversal view of the T1 VIBE Dixon in vivo MRI showing the pancreas with a suspicious lymph node located dorsal from the pancreatic head in the lined box (a). After pancreaticoduodenctomy the fresh resection specimen was pinned on an anatomical drawing (b) and an ex vivo MRI scan was performed on which the same lymph node could be identi ed on the lipid selective image (c) and the water selective image (d). The USPIO-surgery interval was 8 days, thus USPIO was no longer present in the resection specimen and therefore did not elicit any signal intensity changes, After the ex vivo MRI the specimen was xated and sliced into 5mm thick slices (e). Histopathologic examination with hematoxylin and eosin staining revealed a metastasis within the lymph node (e) Figure 6 Example of the work ow in esophageal cancer. Lymph nodes were identi ed on the in vivo MRI after neoadjuvant chemoradiotherapy (a). Nodes could be matched using corresponding anatomical landmarks to the surgical specimen (b) and ex vivo MRI (c, d). The lipid-selective (c) and water-selective (d) ex vivo MRI guided the pathologist towards the lymph nodes for histopathological analysis of each node with hematoxylin and eosin staining (e). The white dashed lined box on the ex vivo MRI (c, d) and pathology (e) contains a healthy lymph node with USPIO contrast (the USPIOsurgery was 1 day). Therefore, the lymph node has lost is signal intensity on the water selective image Example of the work ow in head-and-neck cancer. A lymph node was located on the in vivo MR image (a, white dashed lined box). After lymphadenectomy, the fresh specimen was pinned to a grid (b) and xated with formalin (c). Of the xated specimen, an ex vivo MRI was performed. The lymph node of interest identi ed on the in vivo MR image was correlated to the ex vivo MR image (d). Since the USPIO-surgery interval was 5 days, and USPIO was no longer present in the resection specimen, there were no changes in signal intensity. For this reason, the lymph node has maintained signal on the water-selective ex vivo MR-image (d). The pathologist was guided by the ex vivo MRI to dissect this node and enclose it separately. Histopathologic examination with hematoxylin and eosin staining revealed a healthy lymph node (e) Figure 8 Example of the set-up for ex vivo MRI using the 7T preclinical MR system. From left to right: a formalin xated surgical rectal cancer specimen is placed on a paper sheet. Next it is rolled and placed in a plastic tube and positioned within the MR system Figure 9 Example of a cervical lymphadenectomy specimen vacuum sealed with the TissueSAFE plus device (Milestone, Sorisole, Italy)