Figure 1. Comparison images of study tumor. A, SWIFT image (sweep imaging with Fourier transform) of the 1 of the specimens, with a resolution of 273 μm, shows a large soft-tissue tumor (asterisks). There is cortical interruption (arrow) and extension of the squamous cell carcinoma to the medullary cavity (arrowheads). Soft-tissue invasion of medullary bone is well demarcated from normal fat-containing bone marrow and trabecular structure of medullary cavity. B-F, Corresponding gross image (B) and histopathologic images (C-F) demonstrate infiltrative invasion of the lingual aspect of the mandibular bone along a 15-mm front to a depth of 2.8 mm by a moderately differentiated squamous cell carcinoma. Surgical margins are clear. All histopathologic images use hematoxylin-eosin stain; original magnifications are ×1 (C), ×2 (D), ×40 (E), and ×100 (F).
Figure 2. Comparison images of study tumor. A-F, SWIFT images (sweep imaging with Fourier transform) of the second specimen, with a resolution of 156 μm. A, Regular SWIFT image. B, SWIFT image with short T2 component suppressed. C, Image of short T2 component (subtracting panel B from panel A). D, SWIFT image with fat on resonance and water suppressed. E, SWIFT image with water on resonance and fat suppressed. F, Both fat and water on focus (adding panel D and panel E). Images show cortical interruption and extension of soft-tissue tumor into the medullary cavity. Gross photograph (G) and histopathologic images (H and I) show a well-differentiated squamous cell carcinoma invading along a front of 6.6 mm through the alveolar ridge of the mandible with an erosive and infiltrative pattern for a depth of 1.5 mm (hematoxylin-eosin stain used for both histopathologic images, original magnifications ×1 [H] and ×10 [I]).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kendi ATK, Khariwala SS, Zhang J, et al. Transformation in Mandibular Imaging With Sweep Imaging With Fourier Transform Magnetic Resonance Imaging. Arch Otolaryngol Head Neck Surg. 2011;137(9):916–919. doi:10.1001/archoto.2011.155
Author Affiliations: Departments of Radiology (Dr Kendi) and Otolaryngology (Drs Khariwala and Yueh), Center for MR Research (Ms Zhang and Drs Idiyatullin, Corum, Michaeli, and Garwood), and Department of Laboratory Medicine and Pathology (Dr Pambuccian), University of Minnesota, Minneapolis.
Objective Current imaging techniques are often suboptimal for the detection of mandibular invasion by squamous cell carcinoma. The aim of this study was to determine the feasibility of a magnetic resonance imaging (MRI)-based technique known as sweep imaging with Fourier transform (SWIFT) to visualize the structural changes of intramandibular anatomy during invasion.
Design Descriptive case study.
Setting Tertiary academic institution.
Patients Patients with oral carcinoma who underwent segmental mandibulectomy.
Interventions Two specimens from each patient were imaged using a 9.4-T Varian MRI system. The SWIFT images were correlated with histologic sections.
Results The SWIFT technique with in vitro specimens produced images with sufficient resolution (156-273 μm) and contrast to allow accurate depiction of tumor invasion of cortical and medullary bone. Both specimens had histopathologic evidence of mandibular invasion with tumor. A high degree of correlation was found between magnetic resonance images and histopathologic findings.
Conclusions The SWIFT MRI offers 3-dimensional assessment of cortical and medullary bone in fine detail and excellent qualitative agreement with histopathologic findings. Imaging with the SWIFT MRI technique demonstrates great potential to identify mandibular invasion by oral carcinoma.
Advanced squamous cell carcinoma arising in the oral cavity often invades the mandible. Depending on the degree of invasion, patients may be surgically treated with a marginal or segmental mandibulectomy. In many cases, however, the periosteal layer serves as an adequate barrier to tumor invasion such that mandibulectomy is not required. Unfortunately, detecting bone invasion prior to surgery is often difficult usingcurrently available imaging techniques. The preoperative determination of mandibular invasion with a high degree of accuracy would be useful for several reasons. First, knowledge of invasion necessitating segmental mandibulectomy allows the surgeon to perform the surgery without risking close margins or tumor spill while attempting marginal mandibulectomy. Second, accurate assessment of invasion has the potential to prevent unnecessary mandibulectomy, which can have a significant impact on a patient's functional and cosmetic outcome. Finally, preoperative knowledge of bone invasion allows for accurate planning for the most appropriate reconstructive technique.
Multiple imaging techniques have been applied to preoperatively assess mandibular involvement. These include pantomography, computed tomography (CT), magnetic resonance imaging (MRI), scintigraphy, single photon–emission computed tomography, and 18F-fluorodeoxyglucose positron-emission tomography (PET)/CT.1-8 The most commonly used imaging techniques to evaluate mandibular invasion in oral carcinoma are CT and MRI.
Computed tomography is known to provide high sensitivity and specificity with the application of proper settings (≤3-mm section thickness, isotropic voxel acquisition).2,9 The addition of software applications, such as dentascan, also adds diagnostic value.2,10,11 One of the drawbacks of dentascan is the difficulty in resolving the difference between cortical irregularity and tumor invasion.11 Furthermore, CT is compromised by beam-hardening artifacts (produced by dental amalgam or prosthetic implants), which significantly degrade the assessment of mandibular infiltration by tumor.2,12,13
Magnetic resonance imaging is regarded as a highly accurate examination in the assessment of extension of tumor in the soft tissues, but its role in depicting extension of tumor to the mandible is considered limited owing to frequent overestimation of cortical invasion.3,14,15 This overestimation results from signal changes associated with inflammatory conditions including periodontal disease and peritumoral edema, which are comparable to those seen with neoplastic invasion.2
Like many of the tissues of the musculoskeletal system, cortical bone produces no signal with conventional MR techniques, limiting the characterization of image contrast and differentiation of adjacent soft tissues.16-18 In biological tissues, MR signal comes from the spinning of magnetic moments of hydrogen nuclei. The signal is detectable after a radiofrequency (RF) pulse application. Because the molecular motion within densely mineralized bone is highly restricted, the signal from bone quickly decays after RF excitation. The time constant describing the signal's decay, known as the transverse relaxation time (T2),19 is approximately 200 microseconds in cortical bone. In conventional MRI, excitation and acquisition events are separated by the length of time known as the echo time (TE), which is typically longer than 1 millisecond, which is too long to detect the signal from cortical bone.19-21
The most important factor in mandibular management is to accurately identify the presence of neoplastic invasion and its precise extent by means of preoperative evaluation to obtain adequate oncologic control and to minimize functional sequelae.2 The ideal diagnostic tool should provide noninvasive demonstration of neoplastic invasion of both cortical and medullary bone to allow selection of an accurate surgical strategy.2
A fast and quiet method of MRI known as SWIFT (sweep imaging with Fourier transformation) creates new opportunities for imaging in medicine.19-21 SWIFT uses time-shared excitation and signal acquisition. This allows the detection of signals with a broad distribution of relaxation times, including extremely short T2. This technique offers delineated assessment of cortical and medullary bone, which is not possible with conventional imaging techniques. The present study was designed to assess the feasibility of the SWIFT technique to visualize intramandibular anatomy and potential bony invasion by oral squamous cell carcinoma through comparison with histopathologic findings.
This study was approved by the institutional review board. Two mandibular specimens were obtained after segmental resections. The operative specimens were handed directly to study staff members, who transported the specimens for ex vivo imaging at the Center for Magnetic Resonance Research at the University of Minnesota. Following imaging, the specimens were transported back to surgical pathology for routine histologic processing.
Experiments were performed with a 9.4-T, 31-cm horizontal MRI scanner (Magnex Scientific, Yarnton, Oxford, England) equipped with 30-G/cm gradients (11-cm inner diameter, 300-microsecond rise time) (Magnex Scientific) and driven by a Varian/Agilent Direct Drive console (Varian Inc, Walnut Creek, California). Radiofrequency transmission and signal reception were performed with a home-built, single-loop, 25 mm–diameter coil.
In the SWIFT sequence the excitation bandwidth and acquisition spectral width were both 125 kHz. The repetition time (TR), including 2-millisecond hyperbolic secant pulse length, was 2.5 milliseconds. The total number of projections was 128 000. The average acquisition time was approximately 8 minutes.
Figure 1A and Figure 2A show SWIFT images obtained with the pulse excitation coincident with the tissue water resonance frequency. In Figure 2B and E, A frequency selective pulse was placed +1.5 kHz and −1.5 off resonance from water excitation pulse to suppress short T2 and fat signals, respectively. In Figure 2D, the water signal was suppressed. In Figure 2A, the suppression pulse was replaced by an equal duration delay to make the total acquisition time the same in all Figures. In Figure 2C, subtraction of the SWIFT image with and without the off-resonance suppression pulse provides images of the short T2 component.
Mandibular cortical invasion was diagnosed by interruption or lack of the typical hypointense signal of cortical bone. Mandibular bone-marrow involvement was diagnosed by extension of the soft-tissue tumor to the marrow cavity with replacement of the marrow fat.
Mandibulectomy specimens were grossly examined by the pathologist to assess the presence, size, and location of the tumor and its relationship to the mandible and distance from margins. The specimen was photographed and soft-tissue margins were submitted for histologic examination. The specimen was then fixed in toto in 10% buffered formalin for 24 hours and then decalcified for 1 to 4 days until soft enough to be sectioned with the knife. After decalcification, the mandible together with the surrounding soft tissues and tumor were sectioned at 0.3-cm intervals. The slices were photographed and then processed for paraffin embedding; a single 5-μm section per slice was made and stained with hematoxylin-eosin stain. The presence of bone invasion, site(s) of invasion, pattern of invasion (erosive or infiltrative),22 and the depth of invasion into bone were evaluated by the pathologist (S.E.P.). Histologic sections were scanned on an Aperio CS whole slide scanner (Aperio Technologies Inc, Vista, California), and exact measurements of depth of invasion and distance to margins were made using the digital images. Gross and microscopic images showing the relationship of tumor to bone in the areas imaged by MRI were selected and compared with the MR images.
SWIFT images of the specimens revealed detailed bone and soft-tissue anatomy, including soft-tissue tumor, cortical bone, and medullary bone. SWIFT demonstrated fine anatomic details including nutrient vessels and fine trabecular bone structure. SWIFT produced evidence of cortical bone invasion as cortical interruption of the hypointense signal of cortical bone. Medullary bone invasion was also demonstrated in both specimens as extension of soft-tissue tumor into the medullary cavity and replacement of medullary fat with tumor. The demarcation between tumor-free medullary bone and tumor-invaded medullary cavity is delineated in fine detail (Figure 1 and Figure 2). Short T2 images (Figure 2C) acquired from subtraction of the initially saturated short T2 component from the routine SWIFT image revealed a well-delineated contour of cortical bone.
Histopathologic specimen findings highly correlated with SWIFT imaging findings (Figure 1 and Figure 2).
This preliminary report demonstrates that the SWIFT imaging technique has the capacity to show fine details of intramandibular anatomy. Furthermore, the correlations between the histologic and MR images of these 2 specimens clearly show malignant invasion that has not been previously demonstrated with MR techniques. The data described in this report suggest that MRI has a great deal of potential in accurately determining bone invasion preoperatively.
The main advantage of SWIFT originates in its nearly simultaneous excitation and acquisition scheme. SWIFT allows a TE of almost 0, meaning signal acquisition can begin within a few microseconds after excitation. Thus, SWIFT obtains signal from cortical bone that has a fast-decaying signal, produces less distortion from magnetic susceptibility, and is less sensitive to motion artifacts.18 Also, owing to an incrementally changed gradient, the SWIFT method is nearly 50 dB quieter than a comparable MRI examination.18 In this study we obtained a short T2 image by subtracting an initially saturated short T2 component from the SWIFT image (Figure 2C). Since conventional MRI techniques are not able to fully eliminate the long T2 component, there is always a possibility of false-positive results of cortical bone invasion owing to periodontal disease or inflammation. We believe that the SWIFT technique and the associated short T2 images will overcome false-positive results of cortical bone invasion.
There were limitations to our study. First, we studied only 2 specimens. These specimens were from patients with preoperative clinical and radiologic evidence of cortical and medullary bone invasion. The images served primarily to demonstrate the feasibility of SWIFT to show intramandibular anatomic details. Hence, the strength of the SWIFT technique in assessment of early cortical bone invasion by oral cancer was not evaluated in this study. Nevertheless, given the high-quality images obtained, we are optimistic that this technique will allow identification of very early bone invasion that is not otherwise evident. Third, we are aware that the acquisition of images of similar quality in an in vivo setting with lower magnetic field strength is more challenging. For this reason, the clinical utility of the SWIFT technique needs to be determined in future clinical studies.
In conclusion, our study is very promising in that it offers a SWIFT-based MRI technique for accurate assessment of minute changes of cortical and medullary bone in 3 dimensions without any ionizing radiation. It has the potential to precisely determine the extent of mandibular bone invasion associated with oral carcinoma. This study is a crucial step toward the goal of developing a robust and noninvasive approach for preoperative imaging of mandibular invasion. We are currently in the process of developing a human subject study of SWIFT techniques in a 4T human magnet to solidify the role of this new and transformative technology in the assessment of head and neck tumors.
Correspondence: Ayse Tuba Karagulle Kendi, MD, Department of Radiology, University of Minnesota, MMC 292 Mayo, 8292A, 420 Delaware St SE, Minneapolis, MN 55455 (email@example.com).
Submitted for Publication: May 18, 2011; accepted July 11, 2011.
Author Contributions: Dr Kendi had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Kendi, Khariwala, Zhang, Corum, Michaeli, Pambuccian, Garwood, and Yueh. Acquisition of data: Kendi, Khariwala, Zhang, Idiyatullin, Corum, Michaeli, Pambuccian, and Yueh. Analysis and interpretation of data: Kendi, Khariwala, Zhang, Idiyatullin, Corum, Michaeli, and Pambuccian. Drafting of the manuscript: Kendi, Khariwala, Zhang, Michaeli, and Pambuccian. Critical revision of the manuscript for important intellectual content: Kendi, Khariwala, Zhang, Idiyatullin, Corum, Michaeli, Pambuccian, Garwood, and Yueh. Obtained funding: Garwood. Administrative, technical, and material support: Corum and Yueh. Study supervision: Kendi, Khariwala, Idiyatullin, and Garwood.
Financial Disclosure: Dr Idiyatullin is a consultant for SSI (Steady State Imaging LLC, Minneapolis, Minnesota); Dr Garwood has an equity interest in SSI; and Drs Idiyatullin, Corum, and Garwood are entitled to sales royalties through the University of Minnesota for products related to the research described in this article. These relationships have been reviewed and managed by the University of Minnesota in accordance with its conflict of interest policies.
Additional Contributions: The authors thank Andrew S. Wallschlaeger, PA, for his help with gross specimens, Carrie Zine, HTL (ASCP), for her help with sectioning the specimens, and Jonathan Henriksen for his help with digital gross and microscopic pathology images.