Detecting bone invasion in oral cancer is crucial for therapy planning and the prognosis. The present study evaluated cone beam computed tomography (CBCT) for detecting bone invasion in comparison to standard imaging techniques. A total of 197 patients with diagnoses of oral cancer underwent CBCT as part of preoperative staging between January 2007 and April 2013. The sensitivity, specificity, and accuracy of CBCT were compared with panoramic radiography (PR), multi-slice computed tomography (CT) or magnetic resonance imaging (MRI), and bone scintigraphy (BS) using McNemar’s test. Histopathology and clinical follow-up served as references for the presence of bone invasion. CBCT and BS (84.8% and 89.3%, respectively), as well as CBCT and CT/MRI (83.2%), showed comparable accuracy ( P = 0.188 and P = 0.771). CBCT was significantly superior to PR, which was reconstructed based on a CBCT dataset (74.1%, P = 0.002). In detecting bone invasion, CBCT was significantly more accurate than PR and was comparable to BS and CT/MRI. However, each method has certain advantages, and the best combination of imaging methods must be evaluated in prospective clinic trials.
In addition to the detection of lymph node and distant metastases, the precise determination of the extent of the local tumour in oral cancer is crucial for therapy planning and prognostic stratification. Bone invasion is present in 12–56% of all oral cancer patients. Although the prognostic value of bone infiltration is still controversial, it is highly relevant for the planning of adequate therapy. The suspicion of tumour invasion into the adjacent bone results in radical surgery with wide resection and even microvascular reconstructive surgery, which can significantly reduce the quality of life and increase morbidity and mortality.
Standard preoperative staging consists of clinical examination and non-invasive imaging, including computed tomography (CT) or magnetic resonance imaging (MRI), to detect loco-regional metastasis and to determine the extent of the primary tumour and infiltration into adjacent structures. Previous studies have reported high specificity and low sensitivity for CT and MRI in detecting bone invasion. Bone scintigraphy (BS), which visualizes the bone metabolism of the whole body, is a highly sensitive imaging method. However, given the low prevalence of distant bone metastases in only 2–4% of patients with oral cancer, whole-body imaging seems less appropriate in this regard. Van Cann et al. reported an accuracy of 85% with no false-negative results using CT or MRI in combination with BS for detecting bone invasion. No single imaging modality is considered sufficiently accurate to replace the others.
Over the last few years, cone beam computed tomography (CBCT) has become a routine clinical practice in the three-dimensional (3D) examination of the oromaxillofacial region and in preoperative planning. With this technique, both panoramic radiography (PR) and volumetric datasets, similar to multi-slice CT, can be reconstructed, providing multiplanar and 3D images of the viscerocranium. Compared to conventional CT, CBCT has a lower radiation dose, a higher spatial resolution, and similar or even fewer metal-induced artefacts. However, CBCT is not suitable for assessing soft tissue structures. In our department, CBCT was established in 2007 for diagnostic imaging in prosthetic surgery, implantology, and traumatology in the oromaxillofacial region.
Until now, the role of CBCT in the preoperative assessment of bone invasion in oral cancer has been evaluated only in a few studies with small numbers of patients ( Table 1 ). Comparisons of technologies for non-invasive diagnostic imaging (PR, CT, MRI, and BS) have differed among the published studies. Also, the reported data for CBCT have shown broad variation, with specificity values of 62–100% and accuracy values of 77–96%.
|Patient cohort, n||Imaging methods||Sensitivity||Specificity||Accuracy|
|Hakim et al., 2014||48 a||CBCT||93%||62%||77%|
|Dreiseidler et al., 2011||77||CBCT||92%||96.5%||93.1%|
|Hendrikx et al., 2010||23||PR||54.4%||91.7%||73.9%|
|Momin et al., 2009||50||PR||73% b /56% c||60%||–|
|CBCT||89% b /99% c||60%||–|
|Closmann and Schmidt, 2007||3||Case description|
In this retrospective study, we reviewed a cohort of 197 patients with confirmed diagnoses of oral cancer. To our knowledge, this is the largest cohort that has been used to compare CBCT to other imaging technologies (PR, CT, MRI, and BS) in predicting bone invasion.
Materials and methods
Between January 2007 and April 2013, 352 patients were referred to our department with a suspected diagnosis of oral cancer ( Fig. 1 ). According to our routine staging protocol and current guidelines, a CT or MRI was performed in all of the patients for the assessment of cervical lymph nodes and the extent of the local tumour. A total of 197 of these 352 patients underwent additional BS and CBCT and were therefore included in the study. Medical history, tumour localization, and recent interventions in the oral cavity, e.g., tooth extraction or probe sampling, were documented and accounted for in image interpretation. The imaging findings were validated by histopathology after either a rim or segmental bone resection or a clinical follow-up of at least 6 months ( Fig. 1 ).
The present study was approved by the local ethics committee; no specific informed consent was provided by the patients due to the retrospective nature of the study.
Panoramic radiography and cone beam computed tomography
CBCT was performed using a Galileos CBCT unit (Sirona Dental Systems Inc., Bensheim, Hessen, Germany). The X-ray generator and the detector are mounted across from each other on a U-arm. Both devices rotate around the seated patient. The position of the patient’s head was predetermined with a chin rest and a dental splint. A light localizer tagged the midsagittal plane, and contact was made with the forehead rest in the standardized positioning procedure.
The CBCT features a field of view (FOV) of 15 cm, resulting in a reconstructed 3D volume of 15 cm × 15 cm × 15 cm. The volume consists, in the standard mode, of 512 × 512 × 512 isotropic voxels, with a resolution of 0.3 mm. The scan time is 14 s. The effective dose of this device is 43–175 μSv (85 kV/5–7 mA). Secondary reconstructions display PR-like views (panoramic) and cross-sectional views. The data were saved in DICOM data file format. A more extensive (technical) description of the system can be found in the literature.
Under standardized conditions (tinted room, EIZO RadiForce GS320 diagnostic monitor), two experienced maxillofacial surgeons evaluated the images in consensus using Sidexis XG software, version 2.56 (Sirona Dental Systems Inc.). The diagnosis was first made on the basis of the reconstructed panoramic view (PR). Subsequently, the reconstructed 3D volume rendering of the dataset was assessed in the axial, coronal, sagittal, and tangential views (the Galileos term for a longitudinal view through the tooth, parallel to the mesiodistal dimensions of the teeth-view orientation). All of the sections/planes were continuously adjustable in zoom, contrast, and brightness. The observers were aware of the tumour localization and recent interventions in the oral cavity, e.g., tooth extraction. The absence or presence of bone invasion was judged. Osseous tumour invasion was considered to be present when at least cortical bone erosion or degradation was observed.
For BS, 714 MBq technetium-99m-3,3-diphosphono-1,2-propanodicarboxylic acid ( 99m Tc-DPD) (range 505–820 MBq) was injected; approximately 3 h later, whole-body scintigraphy with continuous scanning in the anterior and posterior views was performed using a double-head gamma-camera with a speed of 15 cm/min, low-energy high-resolution collimators, and a 15% energy window over the 140-keV photopeak of 99m Tc (e.cam systems; Siemens Healthcare, Erlangen, Germany). Regular separate images of the head in the anterior, posterior, and lateral views were added. In cases of uncertain diagnoses (41.6%, 82/197 patients), additional single photon emission computed tomography (SPECT)/CT studies were added using a hybrid system (Symbia T2; Siemens Healthcare). For image acquisition, 180° per detector (also high-resolution, low-energy collimators), a 3° step, a 128 × 128 matrix, and a frame rate of 20 s per frame were used. Reconstruction was performed by iterative ordered subsets expectation maximization (OSEM; 0 Gaussian, 4 subsets, and 10 iterations), both with and without CT-based attenuation correction. A slice step of 5 mm and a slice time of 0.8 s were used to acquire low-dose CT, with CARE Dose modulation and 130 kV over 220°. The transmission data were reconstructed using a B08s kernel to produce cross-sectional attenuation maps, in which each pixel represented the attenuation of the scanned tissue.
All of the bone scintigrams were also evaluated by two experienced nuclear physicians. For image evaluation, the tumour localization and the patient’s history were provided. To obtain a first overview, plain images were initially reviewed in printed form. For the SPECT/CT evaluation, transmission and emission images were fused on a dedicated nuclear medicine workstation (e.soft; Siemens Healthcare) using the software provided. This workstation was also used to evaluate plain images, SPECT, and CT on dedicated monitors.
Bone tissue invasion was suspected if (focally) increased radiotracer uptake in comparison to the surrounding bone tissue or the contralateral bone structures was evident and was observed adjacent to the primary cancer.
Computed tomography and magnetic resonance imaging
For CT, a 64-slice scanner (Somatom Sensation 64; Siemens Healthcare) or a 16-slice scanner (Somatom Sensation 16; Siemens Healthcare) was used. For better soft tissue contrast, the intravenous iodine contrast agent iomeprol (Imeron 300; Bracco Imaging Group, Konstanz, Germany) was administered. Axial and coronal images were reconstructed with a slice thickness of 5 mm in soft tissue and a thickness ranging from 1.25 mm to 5 mm in bone windows.
MRI images were obtained using a 1.5-T scanner (MAGNETOM Avanto; Siemens Healthcare) or a 3.0-T scanner (MAGNETOM Skyra; Siemens Healthcare) before and after the gadolinium contrast agent was administered (gadobutrol; Bayer HealthCare, Germany). T2 fat-saturated STIR (short tau inversion recovery), diffusion-weighted images (DWI), and contrast-enhanced T1 fat-saturated images in different orientations were performed routinely.
On MRI, the tumour signal is quite variable. The purpose of the different weighted images is to detect the signal differences of malignant tissue compared with the surrounding healthy tissue. For example, tumour necrosis often produces a fluid signal that is hypointense on T1-weighted images and hyperintense on T2-weighted images. Many head and neck neoplasms are isointense (have the same signal) to healthy soft tissues (e.g., muscle) on non-contrast-enhanced T1-weighted images and are isointense or hyperintense (have a brighter signal) on T2-weighted images. After the intravenous injection of gadolinium, head and neck tumours might demarcate by enhancing more contrast compared with the surrounding tissue. A T2-weighted, fat-saturated STIR sequence in coronal orientation eliminates the fat signal by turning it completely black and is, therefore, superior to standard T2 sequences. DWI can be used to evaluate the rate of microscopic water diffusion within tissues. Because of the higher density of cells in malignant tumours, the diffusion is restricted compared with the surrounding tissue, which results in a delimitation of the head and neck tumour. Contrast agents could not be applied in patients with reduced kidney function.
The CT and MRI scans were evaluated by two radiologists to assess the extent of the local tumour and cervical lymph node metastasis. On CT imaging, bone tissue infiltration was suspected if there was any alteration in the adjacent bone, including cortical erosion. The images were reconstructed in axial, coronal, and sagittal orientations.
On MRI, cortical bone invasion was suspected in the absence of the typical hypointense signal of the cortex bone on T1- or T2-weighted images. Bone marrow involvement was indicated by a hypointense signal on T1, a hyperintense signal on T2, and/or the presence of contrast agent enhancement. In addition, bone invasion was assumed when a diffusion restriction with signal increase was observed on the diffusion-weighted image and/or a decrease in the apparent diffusion coefficient value was observed.
CT and MRI were considered equal in the determination of bone invasion, in accordance with the guidelines and present literature. If both CT and MRI were performed in the same patient (10/197 patients; 5.1%), the presence of bone invasion was primarily assessed using CT.
Following surgical resection, the tissue samples were submitted to the Institute of Pathology for further macroscopic and microscopic examination, especially to determine the tumour stage and grade ( Fig. 1 ). After appropriate sampling, the tissue was fixed in 4% neutral buffered formalin, and samples containing bony tissue were subsequently decalcified for at least 72 h in formic acid (Merck, Darmstadt, Germany). After paraffin-embedding, 2-μm sections were cut, stained with haematoxylin/eosin (HE), and evaluated by an experienced pathologist.
The results of the different imaging modalities were compared with the histopathology or follow-up data results. The sensitivity, specificity, accuracy, negative and positive predictive values, and false-positive and false-negative values were obtained for each imaging modality. The results of the imaging methods were compared using McNemar’s test. The statistical analysis was performed using IBM SPSS Statistics, version 22.0 software (IBM Corp., Armonk, NY, USA).
The mean age of the patients ( N = 197) at the time of surgery was 63.7 years (range 40–92 years; standard deviation (SD) ±8.8 years). The imaging findings were validated by histopathology in 114 of 197 patients (57.9%), confirming bone invasion in 66 (57.9%) and excluding it in 48 (42.1%).
Due to the absence of clinical signs of bone invasion in the remaining 83/197 patients (42.1%), no bone resection was performed. In these cases, at least 6 months of follow-up (mean 22.3 months, range 6.0–66.3 months, SD ±14.4 months) confirmed the absence of bone invasion, resulting in an overall presence of bone invasion in 33.5% (66/197) of the patients. Further detailed patient information is listed in Table 2 .
|Mean ± SD||63.7 ± 8.8|
|Sex, n (%)|
|Tumour localization, n (%)|
|Soft palate/tonsil||8 (4.1)|
|Floor of the mouth||50 (25.4)|
|Salivary glands||1 (0.5)|
|Tumour stage, n (%)|
|Tumour diameter, cm|
|Mean (range)||2.7 (0.2–8.5)|
|Histopathology, n (%)|
|Adenoid cystic carcinoma||1 (0.5)|